2015 articulation and finger forces in saxophone and clarinet playing (phd thesis) by Adolphesax.com

Articulation and Finger Forces in Saxophone and Clarinet playing Am Institut fĂźr Wiener Klangstil der UniversitĂ¤t fĂźr Musik und darstellende Kunst Wien zur Erlangung des akademischen Grades eines Ph.D. (Doctor of Philosophy) im Fachgebiet der Musikalischen Akustik eingereichte Dissertation von Alex Michael Hofmann

Datum der Einreichung (submitted): Datum der Disputation (defended): Datum der VerĂś entlichung (published):

Begutachter: Ao. Univ.-Prof. Dr. Wilfried Kausel Univ.-Prof. Dr. Christoph Reuter

29. 1. 2015 11. 6. 2015 17. 6. 2015

Abstract Professional woodwind players practise for years until they are able to produce virtuosic performances. While performing, they have to control various parameters at the same time, e.g., ngerings, blowing, lip position and tongue articulation.

This thesis puts a focus on tongue articulation measurements

and nger force measurements in saxophone and clarinet performance. Three empirical studies as well as physical modelling sound synthesis address the question of how players control the instrument to produce expressive tone transitions. First, in a sound production experiment on the saxophone and the clarinet, the participants had to perform melodies which require di erent techniques to play the tones (tongue-only actions, nger-only actions, combined tongue and nger actions) in three tempo conditions. A strain gauge sensor-equipped reed was used to investigate tongue actions of the players during performance. Timing analysis of the captured data showed that in the slow tempo, a combination of tongue and nger actions improved the timing. However, in the fast tempo condition, the timing precision of combined tongue- nger actions was close to the level of nger-only actions, which suggests that the ngering

technique has a dominant in uence on the timing of saxophone performances and clarinet performances.

Second, nger forces applied to the tone holes of the clarinet were measured. Clarinet students and professional clarinetists performed two tasks (expressive performance task, technical exercise task) on a sensor equipped Viennese Clarinet. Although the individual nger force pro les had overlapping tendencies within the two groups of participants, the expressive performance task showed higher nger forces. For the technical exercise task the mean nger forces were lower. In particular, the group of professional players used the lowest nger forces for the technical exercise task. Third, in a listening experiment, it was questioned whether motor expertise in music performance has an in uence on the ability to discriminate articulation techniques in saxophone sounds.

1 published

in Hofmann and Goebl (2014)

Listeners with di erent expertise in

music making (saxophonists, musicians not playing the saxophone, and nonmusicians) participated in an AB-X listening test. Their task was to discriminate saxophone phrases containing legato, portato and staccato articulation. All participants could easily discriminate between staccato articulation and portato articulation, whereas most errors occurred when the listeners tried to discriminate between legato and portato phrases.

In this case, the group of

saxophonists showed the best results, which indicates that expertise in saxo-

phone playing helped to facilitate the task . Finally, the articulation parameters obtained during the experiments were applied to a physical model in an attempt to simulate tongued and non-tongued tone transitions.

1 published

in Hofmann and Goebl (2014)

Zusammenfassung Saxophonistinnen, Saxophonisten, Klarinettistinnen und Klarinettisten benĂśtigen jahrelanges Training um saubere TonĂźbergĂ¤nge und Ăźssige LĂ¤ufe im ausdrucksstarken Spiel erzeugen zu kĂśnnen. WĂ¤hrend des Spielens haben sie unter anderem die Finger, die Atmung, den Lippendruck und fĂźr bestimmte Artikulationstechniken auch die Zungenbewegungen zu koordinieren. Saubere Artikulation, ohne NebengerĂ¤usche erfordert sehr viel Ă&#x153;bung, da fĂźr einige Artikulationstechniken die Zunge direkt das schwingende Blatt berĂźhrt. Der Fokus dieser Arbeit liegt in der Untersuchung von Zungenartikulation und FingerkrĂ¤ften im Saxophon- und Klarinettenspiel. In einer ersten empirischen Studie spielten Saxophonistinnen, Saxophonisten, Klarinettistinnen und Klarinetisten eine technische Ă&#x153;bung, bei deren Umsetzung verschiedene Spieltechniken koordiniert werden mussten (die Zungenbewegungen allein, die Fingerbewegungen allein, kombinierte Zungen- und Fingerbewegungen). Es wurde ein spezielles Sensoreinfachrohrblatt entwickelt, welches es erlaubt in den Blattschwingungen einen Zungen-Blattkontakt zu erkennen. Die Ergebnisse konnten zeigen, dass in einem langsamen Spieltempo die Kopplung von Zungen- und Fingerbewegungen zu einer verbesserten Rhythmik beitrugen. Allerdings war auch zu beobachten, dass im schnellen Tempo die gekoppelten Bewegungen vorwiegend von den Fingerbewegungen beein usst waren. Somit kann man schlussfolgern, dass, obwohl die Zunge den akustischen Tonbeginn steuert, die Finger einen stĂ¤rkeren Ein uss auf

die SpielprĂ¤zision haben . Eine zweite Studie mit studierenden und professionellen Klarinettenspielerinnen und Klarinettenspielern fokussierte auf die FingerkrĂ¤fte beim SchlieĂżen der TonlĂścher. Eine an der Technnischen UniversitĂ¤t eigens dafĂźr entwickelte Wiener Klarinette mit Kraftsensoren wurde fĂźr diese Messungen verwendet. Obwohl die einzelnen Fingerkraftpro le beider Gruppen groĂże Ă&#x153;berlappungen zeigten, wurde dennoch deutlich, dass ausdrucksstarkes Spielen zu einem erhĂśhten Einsatz von FingerkrĂ¤ften fĂźhrte. Beim Spielen einer technischen Ă&#x153;bung wurden demnach im Mittel geringere FingerkrĂ¤fte gemessen und gerade die Gruppe der professionellen Spielerinnen und Spielern zeigte in diesem Fall den geringsten Fingerkrafteinsatz.

1 verĂś entlicht

in Hofmann und Goebl (2014)

In einem HĂśrexperiment wurde untersucht, ob Expertise im Musizieren auch zu einer verbesserten UnterscheidungsfĂ¤higkeit von verschieden artikulierten TonĂźbergĂ¤ngen fĂźhrt (legato, portato, staccato). Von den drei Teilnehmergruppen (Laien, Musizierende die kein Saxophon spielen und Saxophonspielende) zeigten die Saxophoninstinnen und Saxophonisten die besten Ergebnisse. Das lĂ¤sst die Schlussfolgerung zu, dass bei der Unterscheidung von Artikulation in Saxophonmusik mĂśglicherweise Ă¤hnliche VerarbeitungsvorgĂ¤nge im Gehirn

statt nden, wie bei der Verarbeitung von Sprache . AbschlieĂżend wurden die sich verĂ¤ndernden, physikalischen Parameter, welche eine Artikulation mit Zunge von der Artikulation ohne Zunge unterscheiden auf ein physikalisches Model Ăźbertragen. Durch diese Resynthetisierung der KlĂ¤nge wurde verstĂ¤rkt deutlich, welchen Ein uss die Zungenartikulation auf die TonĂźbergĂ¤nge ausĂźbt. Diese EinschwingvorgĂ¤nge genauer zu studieren, kĂśnnte ein Schwerpunkt fĂźr zukĂźnftige Forschung darstellen.

1 verĂś entlicht

in Hofmann und Goebl (2014)

Acknowledgements

This research was carried out within a funded FWF project on Measurement and analysis of nger forces in clarinet playing (P23248-N24).

The FWF

project was a collaboration between the Institute of Sensor and Actuator Systems at the Vienna University of Technology and the Institute of Music Acoustics (IWK) at the University of Music and Performing Arts Vienna. Parts of this work were additionally funded through the MDW-Doktoratsstipendium. I would like to thank, the leaders of this research project, Dr. Walter Smetana and Dr. Werner Goebl, for giving me the opportunity to work within their research groups. First, I would like to especially give thanks to Dr. Werner Goebl for his immense support and guidance throughout the last three years. Thank you for discussing all my research ideas, the progress of my experiments, and for your encouraging words during lunch and co ee breaks.

Furthermore, I am very

grateful to my supervisors Prof. Dr. Wilfried Kausel and Prof. Dr. Christoph Reuter for supporting my research ideas and giving me helpful advise, especially with measurement techniques and signal processing methods. I am also thankful to Dr. Vasileios Chatziioannou for the inspiring collaboration when it came to physical modelling of articulation techniques and for the perfect working atmosphere while we were sharing an o ce. At this point I would like to express my gratitude to Dr. Caroline Cohrdes for her support over all the years and for her indispensable help on statistical methods. The support I have received at the IWK was fantastic.

In particular, I

would like to thank Alexander Mayer for his technical support in the laboratory, Konstantin Zabranski for his support with signal processing, Laura Bishop, Gerald Golka, Michael Weilguni, Saranya Balasubramanian, and Tatjana Statsenko for the inspiring discussions.

Furthermore, I could not have completed this thesis without the help of all the participants who supported my research with their expertise, whether to perform with the sensor instruments or listening to the sound stimuli. Especially, I would like to thank Prof. Oto Vrhovnik, and Dr. Barbara Schickbichler for their support with organising the saxophone study and Dominika Knapp for conducting the listening test. Additional thanks goes to Kerstin HĂśller and Lila Scharang for their helpful advise when preparing the clarinet study and for testing di erent sensor setups on the clarinet. I am also very thankful to Katya Checkovich for proof reading parts of this thesis and for her mental support in the last months of writing. At this place I'd also like to say Thank you!

to all my mentors over

the last decades without their passion for music, their knowledge about music and their patience to teach music, I would not be able to write this thesis. Jan von Klewitz, you ignited my passion for saxophone performance and your encouragement made me nally decide to study music.

Prof. Michael Beil,

thank you for opening my ears to new music, electro-acoustic compositions and new media art. I am also very thankful to Prof. Dr. Stefan Weinzierl and Prof. Douglas Repetto from Technische UniversitĂ¤t Berlin for giving me the opportunity to work as their teaching assistant. Prof. Dr. Herbert Hellhund and Matthias Schubert at HMTM-Hannover, thank you for giving me the chance to develop my personal approach to jazz, live-electronics and saxophone performance and for giving me the freedom to explore music and technology in parallel.

Joachim Heintz from Incontri at

HMTM-Hannover, thank you for being my mentor in computer music, liveelectronics and sound synthesis, thank you for introducing me to the family of Csound experts, and thank you for all your support and especially for being a good friend! Last but not least, I am profoundly grateful to my parents for supporting me over all these years of education.

Contents

Abstract

Zusammenfassung

iii

Acknowledgements

Contents

vii

1 Introduction

1.1

State of the art

. . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2

Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 Measuring musician-instrument interaction on single-reed woodwind instruments 5 2.1

Sound production on single-reed woodwind instruments and the required player actions 2.1.1

2.2

2.3

. . . . . . . . . . . . . . . . . . . . . . .

Investigation of articulatory tongue movements in woodwind performance . . . . . . . . . . . . . . . . . . . . . .

2.1.2

Instrument shapes and ngerings

2.1.3

Tongue and nger coordination in woodwind performance

. . . . . . . . . . . . .

Development of sensor reeds . . . . . . . . . . . . . . . . . . . .

2.2.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . .

2.2.2

Method

. . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2.3

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2.4

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .

Measuring articulation during performance . . . . . . . . . . . .

2.3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.2

Method

. . . . . . . . . . . . . . . . . . . . . . . . . . .

vii

2.4

2.5

2.3.3

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.4

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .

Finger force sensors on saxophone and clarinet . . . . . . . . . .

2.4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . .

2.4.2

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.4.3

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .

General discussion

. . . . . . . . . . . . . . . . . . . . . . . . .

3 Production and perception of legato, portato and staccato articulation in saxophone playing 43 3.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2

Experiment 1: Production task

. . . . . . . . . . . . . . . . . .

3.2.1

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2.2

Results and discussion

. . . . . . . . . . . . . . . . . . .

Experiment 2: Listening test . . . . . . . . . . . . . . . . . . . .

3.3.1

Method

3.3.2

Results and discussion

3.3

3.4

. . . . . . . . . . . . . . . . . . . . . . . . . . .

General discussion

. . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . .

4 Finger forces in clarinet playing

4.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4.2

Performance experiment

. . . . . . . . . . . . . . . . . . . . . .

4.2.1

Methods . . . . . . . . . . . . . . . . . . . . . . . . . . .

4.2.2

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4.3

5 Application of performance measurements to physics based sound synthesis 93 5.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5.2

Physical modelling

5.2.1

Modelling articulation on single-reed woodwind instruments

. . . . . . . . . . . . . . . . . . . . . . . . . . . .

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5.2.2 5.3

. . . . . . . . . . . . . . . . . . . . . . . . .

6 Conclusion and future work

101

6.1

Summary of contributions

. . . . . . . . . . . . . . . . . . . . . 102

6.2

Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

Bibliography

107

A Additional material saxophone experiments

119

B Additional material clarinet experiments

123

Chapter 1

Introduction

1.1

State of the art

Empirical research in music performance has been identi ed as a growing eld in recent decades (Gabrielsson, 2003). There are multiple reasons that explain this immense growth. On one hand, performance research provides detailed insights into the characteristics of a musically professional performance and helps to explain the interpretation process that happens when a professional musician turns a written musical score into pleasant sound events.

On the

other hand, performance research gives the opportunity to understand the complexity of the underlying motor processes humans are able to perform in order to control a musical instrument at a virtuosic level. From the perspective of music acoustic research, the control that a player has over a musical instrument is of particular interest, especially how certain player actions manipulate the resulting sound. The so called Musicianacoustic instrument interaction has been named as a hot topic in recent acoustic conferences (e.g. in Stockholm Music Acoustics Conference in 2013). Expressive performance on musical instruments requires coordinated and goal directed body movements. These body movements have been categorised according to their function in the process of performance (Leman and GodĂ¸y, 2010): Sound-producing gestures, communicative gestures, sound-facilitating gestures and sound-accompanying gestures (Dahl et al., 2009).

Video and

motion capturing technology is useful to observe and categorize player actions at instruments where the player-instrument interaction is visible e.g.,

Chapter 1. Introduction

with drum-set (Dahl, 2004), violin (Schoonderwaldt and Demoucron, 2009; Rasamimanana, 2012), or piano (Goebl and Palmer, 2009, 2008).

Jensenius

et al. (2010) de ned a gesture space for piano performance. He de ned player actions to the keyboard as sound-producing gestured, footpedal actions as

sound-modifying gestured and di erent torso and head movements as ancillary, sound-accompanying, or communicative gestures. Clarinet performances have also been studied with motion capturing technology. Most of these studies focussed on the occurrence and manner of ancil-

lary gestures (Desmet et al., 2012; Caramiaux et al., 2012; Wanderley et al., 2005; Palmer et al., 2009a). One study particularly focused on nger actions in clarinet performance.

Derived from a previous piano study in the same

laboratory (Goebl and Palmer, 2008), Palmer et al. (2009b) showed that clarinetists used larger nger movements and increased accelerations at nger-key contact during di cult and fast musical tone sequences to achieve higher temporal precision. There are still numerous of unanswered questions concerning the ne motor control of ngerings in professional clarinet performance. (e.g., How much nger force do professional players apply to the tone holes? Is there an optimal nger force pro le?) Empirical data of professional clarinetists' nger forces to the tone holes does not exist so far. In this research project an empirical study to investigate nger force under di erent musical situations is foreseen. Weilguni (2013) developed special ring-shaped force sensors for a Viennese Clarinet at the Vienna Technical University. However, most of the sound-producing and sound-modifying player-actions in clarinet (and saxophone) performance happen at the mouthpiece.

Such

embouchure actions include modi cations of the blowing pressure, the lip force, and the lip position (Almeida et al., 2013), but also the control of the players' vocal tract (Scavone et al., 2008), as well as articulatory tongue actions to the single-reed (Liebman, 1989; Wehle, 2007). A focus of this thesis is to measure articulation in clarinet and saxophone performance. Which parameters describe the di erences between articulation techniques (legato, portato, and staccato)? How does the player use the tongue to modulate the reed oscillations directly?

The aim of this research project

is to derive new tone onset parameters (and tone transition parameters) from performance measurements. The obtained parameters will be used to extend

Chapter 1. Introduction

an existing physical model of a single-reed woodwind instrument (Chatziioannou, 2010) towards the simulation of articulation techniques. In woodwind performance, the player has to coordinate articulatory tongue actions and nger actions to produce a sequence of tones. The question, regarding how far these di erent motor processes in uence the quality of the performance is of interest for practising performers and music teachers. Simultaneous measurements of nger actions and tongue actions during performance will provide new insights in sound-producing and sound-modifying gestures on clarinet and saxophone. The results of the experiments are supposed to solve unanswered questions about professional player-instrument interactions.

1.2

Thesis outline

This thesis is structured into four main Chapters (2 5). Each Chapter has its own introduction where the relevant background literature is discussed. This is followed by a report of my research and a discussion of the results. Chapter 2 focusses on player-instrument interactions with single-reed woodwind instruments and discusses performance measurement methods. The proposed measurement techniques put an emphasis on capturing the oscillations of the single-reed, with nger actions at the tone holes simultaneously. Di erent proposals to capture speci c player-instrument interactions are presented and evaluated. Finally, pilot measurements on the saxophone are presented, where nger forces, key accelerations and reed signals are captured simultaneously.

Chapter 3

investigates the production and perception of di erent articula-

tion techniques on the saxophone with two empirical studies. In a production experiment, 19 participants performed two melodies that required di erent motor actions to perform the tones. Timing properties are investigated and compared for the di erent performance conditions. In an adjacent perception experiment, 31 participants with di erent expertise in music making had to discriminate between audio clips with saxophone sounds of common articulation techniques.

The question of whether or not a background in music perfor-

mance has an e ect on the perception of articulation techniques in saxophone sounds is discussed.

1 published

as Hofmann and Goebl (2014)

Chapter 1. Introduction

In Chapter 4, the sensor clarinet which was developed at the Vienna University of Technology (Weilguni, 2013), is used for an empirical study that focusses on nger forces during music performance. In an expressive performance task, excerpts from Clarinet Concerto No.1 in F minor (Op.

73) for

clarinet in Bb, by C.M.v.Weber were performed by 23 participants. In an adjacent technical exercise task, the production experiment from the saxophone study was repeated. Finger force pro les, timing, and articulation techniques of clarinet students and professional clarinetists are analysed in this data. In Chapter 5, the results from the previous experiments (Chapter 2 and 3) are applied to a physical model of saxophone and clarinet. The aim is to enhance the model towards the simulation of tongued and air-separated tone transitions. Hereby, sensor reed signals from di erent performance situations are used to derive meaningful parameter changes, which allow a modelling of these two articulation techniques. The research presented in this Ph.D. thesis was carried out within the FWF project: Measurement and analysis of nger forces in clarinet playing (P23248) at the University of Music and Performing Arts Vienna, under the supervision of the national research partner Werner Goebl. To be in dialogue with the international research community during the project time, preliminary studies were presented and discussed at eligible conferences ranging from the International Symposium on Performance Science to the International

Congress on Acoustics. Reports, which include parts of this thesis, were published in the associated proceedings and one study (Chapter 3) was published as a peer reviewed research article in Frontiers in Psychology. The beginning of each section in this thesis references published articles written by myself, under the supervision of others. The footnotes provide explicit details about my contributions in joint publications.

Chapter 2

Measuring musician-instrument interaction on single-reed woodwind instruments

In this chapter commonalities and di erences between the saxophone and the clarinet are discussed with a focus on the resulting player-instrument interactions. Methods to measure articulatory player actions and nger forces on both instruments are evaluated to prepare the two empirical sound production studies presented later in Chapters 3 and 4.

2.1

Sound production on single-reed woodwind instruments and the required player actions

Clarinet mouthpieces and saxophone mouthpieces share similar construction features.

In both designs, a single reed of cane (or synthetic material) is

mounted to the bottom side of a beak-shaped mouthpiece (Nederveen, 1998; Pinard et al., 2003).

Subsequently, clarinet and saxophone mouthpieces are

controlled with a similar playing technique, where the player encloses the tip of the mouthpiece with his lips, while the front teeth rest on the beak shaped upper side of the mouthpiece. The lower lip covers the lower jaw teeth and gives soft pressure to the reed. The controlling of teeth, lips and jaw is called embouchure (Liebman, 1989).

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

During sound production the player's blowing into the mouthpiece, bends the reed towards the mouthpiece lay. This makes the tip opening smaller, so the amount of air entering the mouthpiece is reduced. At a certain point the pressure that is built up in the instrument turns the reed motion towards the other direction. Constant blowing from the player keeps the reed oscillating, at a frequency which is related to the the impedance peak inside the instrument body (Dalmont et al., 2003; Fletcher, 1979; Almeida et al., 2010). In recent years embouchure related player actions were studied intensively. Gazengel et al. (2007) showed that the lower lip has an important damping e ect on the reed's vibrations.

They showed that without the damping of

the player's lip, the reed's own resonance is signi cantly above the required resonance for sound production on clarinet and saxophone.

Consequently,

wrong adjustments of lower lip pressure from the player in uence the lipdamping and may cause unpleasant squeaks in the sound. Several research teams investigated the in uence of the vocal tract on the sound of the saxophone (Scavone et al., 2008; Guillemain et al., 2010; Chen

et al., 2011). The researchers inserted pressure transducers into the the mouthpiece and into the player's mouth, to measure the impedance spectrum inside the mouth of saxophonists during performance.

These measurements have

shown that in the high register, the vocal tract is crucial in adjusting ne pitch corrections (Scavone et al., 2008). Moreover, Chen et al. (2011) observed that controlling tones in the ageolet register (tones above the standard range of the instrument) was only possible for advanced participants who were able to adjust the resonance of their vocal tract to the targeted pitch.

2.1.1 Investigation of articulatory tongue movements in woodwind performance Tonguing is an important part of musical training and tongue articulation is essential to produce expressive music on woodwind instruments (Bates, 1984; Liebman, 1989; Koch, 1989; Pay, 1995).

Clarinetists and saxophonists use

their tongue to directly modulate the oscillating reed to shape tone onsets and tone o sets. For a long time, the examination of tongue motions during woodwind perfor-

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

mance has been an interest in the teaching of wind instruments. Medical equipment, mostly from speech therapy, has already been used to monitor tongue actions during performance. Gardner (2010) used ultrasonograhic imaging to explain the function and movement trajectories of the tongue in professional clarinet performance.

He observed relatively stable tongue motion patterns

for recorded repetitions of two production tasks: speech production and clarinet articulation under di erent vowel and consonant conditions.

With this

method, it was possible to show that in clarinet performance the tongue shape remained stable across the entire pitch range of the instrument. A limitation of this study was that all the data gained was from one recording session with only one participant (the author himself ). Gardner also stated that even at a slow playing speed, the frame rate of the ultrasound equipment (28 Hz) was not ideal for capturing high speed tongue motions. Consequently, new methods to capture tongue actions during performance are required. A method to capture tongue-reed interactions, using strain gauge sensors on the reed will be discussed later in this chapter.

2.1.2 Instrument shapes and ngerings In literature, the vibrating reed has been described as a mechanical oscillator that acts as a pressure-controlled valve to modulate the ow into the instrument body (Kergomard, 1995). The instrument body functions as a resonator. Its impedance peak is determined by its length and shape. The reed oscillates at a frequency related to this impedance peak. Tone holes ( nger holes or key holes) along the instrument body allow the player to modulate the frequency of the tone. On the saxophone and the clarinet, ngers from both hands are required to operate the tone holes. To play larger intervals, the player has to close or open multiple tone holes at the same time. This requires simultaneous nger movements, also called safe nger transitions. With a measurement on the ute, Almeida et al. (2009) was able to show that unsafe nger transi-

tions in the dimension of a few milliseconds (21 ms) already in uenced the radiated sound. Professional ute players were able to use smaller delay times between their individual ngers compared to novice players. This had a positive in uence on their sound. Virtuous, expressive performance on woodwind

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

instruments requires nger movements with precise timing. Professional musicians have to use their stamina e ciently to be able to play several hours per day and to avoid overexertion (Nemoto and Arino, 2007; Spahn et al., 2011). Not much nger force is required to close a tone hole and the force applied after the tone hole is closed has no in uence on the resulting sound.

Nevertheless, instrument teachers often reported that students use

too much nger force. Music education encourages loose ngering technique (Wehle, 2007).

Finger force measurements on clarinet instruments, foreseen

in this thesis, will gain new insights into the used nger forces of professional players under di erent playing conditions. The ngering schemata of clarinet and saxophone di er because of a main constructional di erence between the instruments. Clarinets have a cylindrical tube (instrument body), whereas saxophones have a conical tube. The reason for this di erence comes from the fact that the size of the tone holes in the instrument body has an in uence on the maximum dynamics of the radiated sound (Hall, 1991). The predominantly cylindrical tube of the clarinet (cylinder with approximately 15.5 mm diameter) allows only small tone holes (3 12 mm). To build a single-reed instrument with larger tone holes, the shape of the tube must be changed. Consequently, a conical tube of the same length, as with the soprano saxophone (9 90 mm) allows larger tone holes (1 31 mm) and can produce louder sounds than a clarinet (Nederveen, 1998). Although both instruments (clarinet and soprano saxophone) share the same excitation mechanism and their tubes are of approximately the same length (650 mm), the shape of the tube di erentiates the two instruments. In the sound of the saxophone, all members of the harmonic series appear, which is a characteristic for the sound. Thus, overblowing to the rst harmonic allows an octave shift of the tones from the lower register. On the saxophone, the same ngerings are used to play in the higher register.

Technically, the

large tone holes can not be covered by the ngers alone. The player has to use the key pads which are located over the large tone holes. Conversely, the sound of the cylindrical shaped clarinet includes only the odd-numbered members of the harmonic series, which produces a di erent timbre.

Furthermore, this a ects overblowing on the clarinet.

Because of

the missing second harmonic, overblowing produces the sound of the third

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

harmonic, a fth beyond an octave. Hence, the clarinet requires at least 19 ngerings to produce all tones in the lower register and needs di erent ngerings for the overblown register. This has consequences on the clarinet key work.

Although, the small tone holes can be closed by the ngers directly,

additional tone holes are required for chromatic playing.

These are coupled

to a complicated key work including rings around the tone holes and keypads. Throughout the history of clarinet making (1770 until the early part of the twentieth century), the clarinet key work was still improved (see Shackleton, 1995, for an overview). As a result, two slightly di erent ngering systems for the clarinet are common today: the French (Boehm) system and the German (Oehler) system. In Austria, where this research project took place the Viennese clarinet is predominantly played. Viennese clarinets share most of the construction features of German clarinets , including the German (Oehler) ngering system. The main di erence is a thicker instrument body with slightly larger tone holes, in comparison to other German clarinets (Birsak, 2009).

2.1.3 Tongue and nger coordination in woodwind performance In saxophone and clarinet playing the ngers and the tongue have to be coordinated.

Depending on the musical situation, the ngers (legato-playing),

the tongue (tone-repetitions), or a combination of both (portato playing and staccato playing) in uence the timing and the sound of the tones. Recent performance studies on wind instruments focused mainly on one single control parameter at a time e.g., the ngering technique (Palmer et al., 2009b), the tonguing technique (Gardner, 2010), the lip pressure (Gazengel

et al., 2007), or the vocal tract in uence on the sound (Scavone et al., 2008). Studies on tongue- nger coordination in woodwind performance do not exist. To gain information about tongue nger-coordination in single-reed woodwind performance, a measurement method that can capture tongue and nger actions simultaneously is required. An important issue is that this technology should not handicap the player, but capture reliable data and be suitable for empirical studies. In this chapter, sensor technology applicable to saxophone

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

and clarinet are discussed, which allow investigations of tonguing technique and ngering technique during performance.

2.2

Development of sensor reeds

This section describes a method to equip synthetic single reeds with sensors in order to measure the oscillations of the reed during performance. Parts of

this section were already published in Hofmann et al. (2013a) . The following section (2.2) is based on this article.

The section also further includes

unpublished measurement results for clarinet reeds.

2.2.1 Introduction2 Saxophone reeds and clarinet reeds have a long tradition to be manufactured from natural cane (Nederveen, 1998; Shackleton, 1995).

The material prop-

erties of cane are well suited for this sound producing application, but the disparity and sensibility of wooden reeds, with the resulting limits in operating time were often criticized (Liebman, 1989). The hygroscopicity of natural cane leads to a variation in the elastic properties of the material.

To be in

playing condition, traditional reeds have to be wet (Pinard et al., 2003). Dryness worsens the sound quality.

Chemists worked on synthesizing materials

which are applicable for a similar application but without these disadvantages. In the following section the properties of synthetic reeds from the company

LĂŠgĂ¨re are discussed, because these reeds were used in this research project. LĂŠgĂ¨re synthetic reeds are made out of an oriented polymer with material properties (density: 0.9 1.1 g/mL; Young's modulus: 5 10 GPa) copied from cane reeds in playing condition (Legere, 2000). The bending sti ness of the material determines the reed strength. The elastic modulus depends on the

1 As

the rst author of this publication, I made the following main contributions: the construction and the design of the sensor reeds, the design of the experiment, the data analysis, and the writing of the research report. Calibration measurements were carried out at the Institute of Sensor and Actuator Systems, Vienna University of Technology, in collaboration with Michael Weilguni and Vasileios Chatziioannou. The procedure of the experiment was under the supervision of Werner Goebl and Wilfried Kausel. 2 as published in Hofmann et al. (2013a)

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

molecular alignment of the polymer and can be regulated by the manufacturer to produce reeds of di erent strength, thus all reeds are based on an identical cut (like with cane reeds). reed's sound properties.

Further e orts were undertaken to improve the

By keeping the polymer's density, while increasing

the Young's modulus (up to 16 GPa) plus making the reed 20% thinner, the manufacturer reported that those reeds' sound was judged to be similar to the sound produced by cane reeds (Legere et al., 2011). Synthetic single-reeds gather several properties which favour them over natural cane reeds, especially for research applications.

Non-hygroscopic reeds

provide better conditions for direct comparisons of measurements from di erent time. Synthetic reeds are easy to clean and thereby suitable for empirical studies. Besides, they have a longer lifetime, because the material is designed to avoid splitting at the tip. Furthermore, lost or destroyed reeds can be replaced by replicates during an experiment because the material properties are de ned by the manufacturing method.

2.2.2 Method1 Strain gauge sensors Bonded resistance strain gauges are based on the principle that the resistance of an electrical conductor changes when it becomes stretched. Foil strain gauges are produced by a low cost circuit printing technique and function as a stable, accurate, self temperature compensated measurement instrument. foil, with polyimide backing, allows a measurement lifetime of more than

The

105

cycles (Window, 1992). In this project, the sensor is used in a small operation temperature range (20 36

â&#x2014;Ś

C), but in a wet environment. Hence, sensor

and wires have to be protected against moisture. The sensing area of a strain gauge is a small (copper nickel alloy) grid on the foil.

The sensor averages

the strain over this area by changing its resistance when bent. Consequently a deformation of the sensor covered area of an object can be measured.

capture the sensor's resistance several Wheatstone bridge circuit designs are available for disposal (Window, 1992). In the current setup a quarter bridge circuit, with 5V power supply and additional signal ampli cation (INA 126,

1 as

published in Hofmann et

al.

(2013a)

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

70 mm

65 mm

ﬂat side

6 mm

26 mm

12 mm

16.3 mm

4 mm

30 mm

curved side

ﬂat side ﬂat side

sensor area 3.1 mm

sensor area

0.2 mm

3.1 mm

0.2 mm

curved side

Figure 2.1: Schematic of an alto-saxophone reed (left) and a Bb clarinet reed, german cut (right), equipped with a 2 mm strain gauge.

Figure 2.2: Synthetic alto-saxophone reed (left) and synthetic Bb clarinet reed, german cut (right), equipped with a 2 mm strain gauge. by Texas Instruments Incorp.) was used.

Reed preparation Standard industry foil strain gauge sensors (RS, 2 mm, 120 ohms) were attached to the at side of a synthetic reed (see Figure 2.2).

Attaching the

sensor to the side of the reed which shows inside the mouthpiece ensures that during playing the player's lip does not touch the sensor.

The sensor was

placed with some distance from the tip to avoid preparations at the thinnest and most sensitive area of the reed. The reeds were modi ed by the following four steps. First, the strain gauge sensor was glued on the at side of the synthetic reed, with a distance of 4 mm (alto-saxophone reed) to the tip. Second, two

0.5 mm

holes were drilled

near the heart of the reed, to lead the cables to the other side of the reed, where a connection socket can be mounted. Third, the sensor and the cables were protected with lacquer (transparent nail polish) against moisture inside the mouthpiece chamber. to make them airtight.

Finally the drill holes were lled with hot-glue,

The design properties had to be adjusted for the

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Figure 2.3: Left: Clarinetist's embouchure while playing on a Bb clarinet. Right: Clarinetist's embouchure on a Bb clarinet, sensor reed connector pins placed between lips and ligature on the mouthpiece.

particular dimensions of Bb- clarinet and alto-saxophone reeds. See Figure 2.1 for respective dimensions.

Sensor calibration measurements A universal testing machine (Inspekt micro LC 100N by Hegewald & Peschke, Germany) was used to characterize the properties of a sensor equipped altosaxophone reed (LĂŠgĂ¨re, strength: 2.25) and a sensor equipped Bb-clarinet reed (LĂŠgĂ¨re, German cut, strength: 4.25). For the alto-saxophone, the sensor reed was mounted to a Vandoren (AL3, 1.52 mm tip opening) mouthpiece through the associated Vandoren Optimum ligature. A customized mounting device held the mouthpiece tight to a motor driven table of the testing machine (see Figure 2.4). Lifting of the table caused the reed to press against the load cell (measurement head). The machine lifted the table by 20

Âľm , while a s

100 newton load cell tracked forces occurring at its

tip (sampling rate: 10 Hz). During this process the strain gauge sensor signal was captured simultaneously. For measurements with the clarinet sensor reed a Maxton mouthpiece (NA-1, 0.75 mm tip opening) was used. The sensor reed properties were measured under three di erent conditions. In the rst condition the reed was bent by the resistance of the load cell, applied to the tip of the reed (20

Âľm for s

1500 Âľm,

Figure 2.5). In the second

measurement condition, a soft clamp was added to the setup to the position of the player's lip (14 mm from the tip), to simulate an embouchure (Figure 2.6).

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Figure 2.4: Clarinet mouthpiece with sensor reed, mounted to a universal testing machine (Inspekt micro LC 100N by Hegewald & Peschke, Germany). Setup for measurement condition one, bending the tip of the reed.

Again, the reed was bent by the resistance of the load cell, applied to the tip of the reed. In the third measurement condition the load cell was applied to the position of the player's lip, in an attempt to measure e ects of the player's lip to the sensor signal (Figure 2.7).

2.2.3 Results1 For all measurements, the displacement of the table, the measured voltages from the strain gauge sensor circuit and the force at the load cell were captured simultaneously, while the machine lifted the table.

1 as

published in Hofmann et

al.

(2013a)

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Load cell

Displacement

Strain gauge

0.0

0.5

1.0

1.5

2.0

1.5 1.0

Reed displacement (mm) Strain gauge signal (V)

0.5

1.0

Reed displacement (mm) Strain gauge signal (V)

0.0

Displacement (mm) / Strain gauge signal (V)

1.5

Clarinet Reed

0.0

Displacement (mm) / Atrain gauge signal (V)

Saxophone Reed

0.0

0.5

Force (N)

1.0

1.5

2.0

Force (N)

Figure 2.5: Measurement Condition 1: The load cell's (measurement head) resistance bends the tip of the reed (top). Left: Force-displacement curve (solid, black line) and strain gauge sensor signal (dashed, red line) for the alto-saxophone reed (strength: 2.25). Right: Force-displacement curve (solid, black line) and strain gauge sensor signal (dashed, red line) for the clarinet reed (strength: 4.25).

Measurement at reed tip In the rst measurement condition, the reed was bent through the resistance of the load cell at the tip of the reed. Figure 2.5 shows the measurement setup and the two graphs with the captured data for both reeds. The solid black line depicts the displacement of the reed's tip in relation to the force at the load cell, while the dashed red line depicts the measured voltage deviations of the strain gauge sensor. The diagrams in Figure 2.5 show that the measurements of the reed displacement and the strain gauge signal overlap, until a certain point, where the reed closes the mouthpiece completely (Force with approximately 1 N).

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

At this point a deformation of the reed was visually observed during the experiment.

The measurements show a linear reed displacement curve, until

the closing point. This linear trend was opposite from our expectations. We expected a non linear force-displacement curve, representing the entire reedmouthpiece system, where the bending sti ness is raised by the reed touching the mouthpiece lay (beating, Dalmont et al., 2003). Instead the results of our measurement show a force pro le, similar to the spring constant of a free reed. This can be explained by the shape of the load cell. In our measurements the head acted as a pointing force to only the tip of the reed. As a result only the tip was bend and the reed did not touch the lay. Accordingly, the measurement shows the local spring constant of the tip of the reed. In a real playing situation the player's embouchure bends the reed towards the mouthpiece lay and the Bernoulli force from the oating air stream, acts on the whole reed area. Under these conditions the reed would beat against the mouthpiece lay. However, Figure 2.5 shows that the strain gauge sensor signal overlaps the force-displacement curve, hence the strain gauge measures the bending of the reed's tip.

Similar observations were made for the clarinet reed (see Figure

2.5, right).

Measurement at reed tip with arti cial lip In the second measurement condition, a soft clamp was added to the position of the player's lip (14 mm from the tip). The lip clamping force was adjusted to that the tip opening was reduced to 0.5 mm for the saxophone and 0.4 mm for the clarinet. The load cell was applied with 3 mm distance from the tip of the reed. The diagrams in Figure 2.6 show that the measurements of reed displacement and strain gauge signal show similar characteristics to the rst measurements. Displacement and strain gauge signal behave linear until the point where the reed closes the mouthpiece tip (Force with approximately 0.75 N). Although this measurement setup is closer to a real playing situation, the pointing force to the tip of the reed does still in uence the measurement results. The measured curve depicts the local spring constant of the tip of the reed without beating the mouthpiece lay.

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Load cell

ArtiďŹ cial lip

Displacement

Strain gauge

0.0

0.5

1.0

1.5

1.5 1.0

Reed displacement (mm) Strain gauge signal (V)

0.5

1.0

Reed displacement (mm) Strain gauge signal (V)

0.0

Displacement (mm) / Strain gauge signal (V)

1.5

Clarinet Reed

0.0

Displacement (mm) / Strain gauge signal (V)

Saxophone Reed

2.0

0.0

Force (N)

0.5

1.0

1.5

2.0

Force (N)

Figure 2.6: Measurement Condition 2: Arti cial lip added to the setup. The load cell (measurement head) bent the tip of the reed (top). Left: Force-displacement curve (solid, black line) and strain gauge sensor signal (dashed, red line) for the altosaxophone reed. Right: Force-displacement curve (solid, black line) and strain gauge sensor signal (dashed, red line) for the clarinet reed.

Measurement at the position of the lip In the third measurement condition, the load cell was applied to the position of the lip (14 mm from the reed tip), to measure the e ect of the player's lip force to the reed and particularly its in uence to the strain gauge sensor signal (Figure 2.7, top). Under this measurement condition a non-linear forcedisplacement curve was observed (Figure 2.7, bottom, solid line). Positioning the load cell to approximately the middle of the free reed part, changed the results of the measurements extremely.

First, the displacement-force curve

shows a non-linear behaviour, which can be explained as the e ective sti ness of the reed-mouthpiece system.

Compared to the spring constant of a free

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Load cell

Displacement

Strain gauge

1.5 1.0

Reed displacement (mm) Strain gauge signal (V)

0.5

1.0

Reed displacement (mm) Strain gauge signal (V)

0.0

Displacement (mm) / Strain gauge signal (V)

1.5

Clarinet Reed

0.0

Displacement (mm) / Strain gauge signal (V)

Saxophone Reed

Force (N)

Figure 2.7: Measurement Condition 3: The load cell (measurement head) bent the reed at the position of the player's lip (top). Left: Force-displacement curve (solid, black line) and strain gauge sensor signal (dashed, red line) for an alto-saxophone reed. Right: Force-displacement curve (solid, black line) and strain gauge sensor signal (dashed, red line) for a clarinet reed.

reed, the bending sti ness is raised by the reed touching the mouthpiece lay (beating Dalmont et al., 2003). Second, Figure 2.7 (bottom, dashed red line) shows that while bending the reed at the position of the player's lip, the strain gauge signal remained constant. This proofs that the sensor reed signal is not a ected by the lip force of the player's embouchure. However, when the reed closed the mouthpiece tip, the deformation of the reed can be seen in the strain gauge signal (Saxophone reed: force = 14 N, displacement = 0.6 mm; Clarinet reed: force = 7 N, displacement = 0.4 mm).

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

2.2.4 Discussion1 This section explained a method to equip synthetic single-reeds with strain gauge sensors in an attempt to measure reed bending during musical performance.

In a material characterization experiment with an universal testing

machine, reed displacement and strain gauge signal (reed bending) were measured, in relation to the applied force. A force applied to the tip of the reed showed that the displacement curve was correlated to the strain gauge signal. Applying the pointing force to the position of the player's lip, revealed no changes in the strain gauge signal. This observation suggests that the strain gauge signal gives useful information about the bending of the tip of the reed. Further investigations may help to relate the reed-bending parameter to the displacement of the reed's tip (tip opening). Tip opening is an essential parameter when working with physical models of single-reed woodwind instruments (Chatziioannou, 2010).

Deriv-

ing reliable tip opening values from reed bending measurements would allow direct comparisons of the reed behaviour in real playing situations with calculations from physical models (Chapter 5). However, calibrating the sensors on the reeds remains a complicated topic.

Reasons for measurement errors

may be the position of the sensor on the reed and the size of the strain gauge sensor in general.

Conclusions about the vibrations of the whole reed and

further insights on the tip opening parameters are limited. Future work foresees investigations with a high speed camera and an arti cial blowing machine to gain additional information to further expose the question of reed sensor calibration.

2.3

Measuring articulation during performance

This section will focus on the measured e ects of articulatory tongue actions to the vibrating single-reed, captured under real playing conditions. Parts of this section have been presented and discussed at two international conferences

and were published in the associated proceedings (Hofmann et al., 2012a

1 as

and

published in Hofmann et al. (2013a) am the rst author of this publication and made the following main contributions: the design of the experiment, the data analysis, and the writing of the research article. The 2I

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Hofmann et al., 2013a). For this section of my thesis, previously unpublished measurements with the clarinet were added.

2.3.1 Introduction 1 Woodwind players may choose from a large repertoire of articulation techniques, that allows them to shape tone onsets and tone o sets according to the intended musical expression. One technique which is often used is portato articulation.

Portato articulation is produced by soft tongue strokes to the

reed, while the player's blowing is constant (Koch, 1989). For fast sequences some professional players also use the technique of double tonguing. Instead of a second tongue stroke to the reed, "the hump portion abruptly rises up striking the roof of the oral cavity which in turn stops the air ow from the larynx" (Liebman, 1989, p. 31). Liebman describes the resulting sound e ect as less e ective on saxophone than on the ute. Di erent attempts to explain the articulatory functions of the player's tongue have been made in the past, in an attempt to develop more realistic models for physics based sound synthesis.

Two contrary explanations of the e ects

of tongued articulation to the vibrating reed have been discussed lately. On the one hand, Ducasse (2003) describes that a tongue stroke to the reed has a damping e ect to the vibrating reed and the force of the tongue changes the equilibrium position of the reed. On the other hand, in the model of Sterling

et al. (2009) the tongue is understood as a gate in front of the mouthpiece tip, which only prevents the air-stream to enter the mouthpiece. This experiment investigates tongued and air-stream separated tone sequences, performed on alto-saxophone and Bb clarinet in an attempt to gain detailed insights into the reed behaviour during such tone transitions.

measurements in the laboratory were carried out in collaboration with Vasileios Chatziioannou. The procedure of the experiment was under the supervision of Werner Goebl, Wilfried Kausel, and Walter Smetana. 1 as published in Hofmann et al. (2012a)

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

2.3.2 Method To examine the e ects of articulation techniques to the sound excitation mechanism on saxophone and clarinet, three control parameters were captured during performance:

the blowing pressure from the player, the bending of the

reed and the resulting inner mouthpiece pressure. The study was undertaken in two parts.

At rst, measurements were done on the saxophone with a

simpler setup . For the second measurement the setup was enhanced. More precise pressure transducers, similar to the ones in (Scavone et al., 2008) were used, to study more details of the embouchure at the clarinet. As the focus of this study are articulation techniques and the resulting reed behaviour, only a single-microphone measurement (inside the mouthpiece) will be used. Other research groups (van Walstijn and de Sanctis, 2014; Chatziioannou, 2010) used multiple microphones to track the wave propagation inside the resonator.

Experimental setup To capture the blowing pressure, the reed bending and the mouthpiece pressure simultaneously during performance, the following setup was prepared for

measurements on the alto-saxophone : First, the blowing pressure of the player was measured by a pressure transducer (Technoterm 5402), which was attached to a probe that ranged into the player's mouth. Second, the reed bending was measured with a synthetic, sensor equipped, saxophone reed as described in Section 2.2. Third, the inner mouthpiece pressure was measured by a small condenser microphone (40DP 26AS, by G.R.A.S.) inserted into the chamber of the mouthpiece (AL3, by Vandoren) through a hole on one side. All three channels were recorded onto computer hard disk (A/D conversion DAQ LabView 2011, by National Instruments Corp.)

using a sampling frequency of

11.025 kHz (16 bits). For this experiment the mouthpiece was only connected to the alto-saxophone neck. For measurements on the Bb-clarinet, an enhanced measurement setup was choosen: Calibrated piezo-resitive pressure transducers were used to measure precise pressure values (in Pascal) inside the mouthpiece and inside the players

1 Hofmann

2 Hofmann

et al. et al.

(2012a, 2013a) (2012a)

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

mouth. The recordings were made with a higher sampling frequency of 44.1 kHz (16 bit). A simple cylindrical tube was used as the resonator, in an attempt to collect data applicable for direct comparisons of the measured signals to the output calculated with physical model simulations (inverse modelling, Chapter

5).

The experimental setup (see Figure 2.8) consisted of a clarinet mouthpiece (Maxton NA-1) attached to a cylindrical tube (length

l = 0.33

m; radius

r = 7.5 mm), two piezo-resitive pressure transducers (Endevco 8507C-2) and a synthetic clarinet reed (LĂŠgĂ¨re, German cut, reed strength 3.75) with a strain gauge sensor attached to it (described in Section 2.2).

Both piezo-resistive

pressure transducers were operated with the accompanying DC di erential voltage ampli er (Endevco Model 136). One pressure transducer was directly inserted into the chamber of the mouthpiece through a drill-hole. The second pressure transducer was connected to a probe, that ranged into the player's mouth.

Figure 2.8: Clarinet mouthpiece with sensor-reed and two pressure transducers, capturing blowing pressure and inner mouthpiece pressure.

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Procedure A professional woodwind player (author) performed two common techniques of tone transitions, with mezzo piano dynamics.

In the rst case (tongue-

separated tones) a portato sequence was performed, where the player used the front area of the tongue to directly interact with the reed to separate the tones (as described in Liebman (1989); recordings see Figure 2.9 Timespan:

3 â&#x2C6;&#x2019; 5 s).

In the second part of experiment (Timespan:

5 â&#x2C6;&#x2019; 7 s)

(air-separated

tones) the same tone sequence was produced, but this time the player followed the instructions for the technique of double tonguing (raise the hump portion of the tongue, to strike the roof of the oral cavity, which stops the air ow from the larynx). Additionally a tone was played, where the blowing was increased slowly, until the sound starts. The experiment was done with the alto-saxophone setup rst and later repeated with the enhanced clarinet measurement setup.

Figure 2.9: Recorded signals on the saxophone, showing the reed bending (top panel), the normalized inner mouthpiece pressure (middel panel) and the normalized blowing pressure (bottom panel) under two playing conditions: Tongue separated tones (3 5 s) with a constant blowing pressure. Air-separated tones (5 7 s), with a varying blowing pressure. The blowing pressure measurements verify the applied playing technique.

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Figure 2.10: Normalized inner mouthpiece pressure (bottom) and reed bending (top) for increased blowing until sound occurs on the saxophone.

2.3.3 Results1 Increasing blowing pressure On the saxophone, a tone was played without using the tongue, by continuously increasing the blowing until the sound starts.

Then the player immediately

stopped blowing. The bending signal (Figure 2.10, top) shows that the blown air bends the reed towards the mouthpiece lay. At a certain point (2.2 s) small oscillations begin and increase, until a steady state condition of the tone is reached. When the player stops to blow, the tone ends and the reed returns to the rest position. The same procedure was recorded with the clarinet setup which captures the pressure values of blowing and resulting pressure in the mouthpiece. Figure 2.11 shows pressure values, measured inside the player's mouth (bottom panel) and inside the mouthpiece (middle panel), in relation to the reed bending signal. At the very beginning (Figure 2.11, 0.0 s 0.4 s), the player removes the tongue from the reed and slowly increases the blowing pressure (0 2 kPa). The increasing air ow bends the reed towards the mouthpiece lay (0.4s 1.5s). At a certain point (1.5 s) the reed begins to oscillate. Reducing the blowing

1 as

published in Hofmann et

al.

(2012a) and Hofmann et

al.

(2013a)

Chapter 2. Measuring musician-instrument interaction on single-reed woodwind instruments

Figure 2.11: Measured reed bending signal (top panel), mouthpiece pressure (middle panel) and blowing pressure (bottom panel). Player removes tongue from the reed and then increases blowing into the clarinet mouthpiece until sound starts.

pressure leads to a release of the tone (3.0 s).

Observed signals during sound production The sound excitation principle on single-reed woodwind instruments can be observed with this measurement technique. Zooming into a steady state part of the sound shows the interaction between reed motion and pressure inside the mouthpiece.

In Figure 2.12, reed bending and mouthpiece pressure are

plotted in one graph. When the player's blowing is bending the reed towards the mouthpiece (begin of red curve), the reed reduces the tip opening. This motion is additionally reinforced by the Bernoulli-force from the ow of air into the mouthpiece. The smaller tip opening lowers the air pressure in the mouthpiece (decreasing blue curve). At one certain point (e.g. 0.002 s), the re ections from inside the resonator (instrument) turn the reed motion to the other direction and open the mouthpiece tip. Constant blowing from the player makes the system oscillate.

Chapter 2. Measuring musician-instrument interaction on single-reed

1.0

woodwind instruments

â&#x2C6;&#x2019;1.0

â&#x2C6;&#x2019;0.5

0.0

0.5

Reed bending Mouthpiece pressure

0.000

0.005

0.010

0.015

0.020

0.025

0.030

Time (s)

Figure 2.12: Normalized inner mouthpiece pressure and reed bending during a steadystate tone on the clarinet.

Articulation Blowing pressure, reed bending and inner mouthpiece pressure were captured simultaneously for two contrasting articulation techniques, respectively tongue-

separated and air-separated tone transitions (Figure 2.13). The blowing pressure signal shows that for tongue-separated tones, the blowing from the player was approximately constant through the note transition (0.2 s, Figure 2.13).

In contrast, to produce air-separated tones the player

changed the blowing pressure during the note transition. The enhanced measurement setup used for the clarinet gives absolute values of blowing pressure at 4.5 kPa during tone production in mezzo piano playing. This value remained constant for tongue-separated tone transitions but dropped down to 0.7 kPa for an air-separated tone transition. The blowing pressure signal additionally veri es that the player followed the instructions for the two di erent playing techniques.

Extracted parameters: Tongued tone transitions When the player strokes the reed with the tongue (tongue-reed contact, TRC) this abruptly bends the reed towards the mouthpiece. The soft tongue surface dampens the oscillations of the reed (Figure 2.13, top left). This leads to a sudden and short reduction of the pressure inside the mouthpiece (Figure 2.13 middle left). When the tongue releases the reed (TRR), the reed bounces back

Chapter 2. Measuring musician-instrument interaction on single-reed

Reed bending 0.2

0.3

2 0 â&#x2C6;&#x2019;4 5 4 3 2 1

Blowing pressure (kPa) 0.1

Mouthpiece pressure (kPa)

4 2 0 â&#x2C6;&#x2019;4 5 4 3 2 1 0 0.0

0.0

0.4 0.0

Reed bending Mouthpiece pressure (kPa) Blowing pressure (kPa)

0.4

b) 0.8

0.8

woodwind instruments

0.4

0.5

Time (s)

0.6

0.7

0.8

0.9

Time (s)

Figure 2.13: Measured reed bending signal (top panel), mouthpiece pressure (middle panel) and blowing pressure (bottom panel) for a single note transition on the clarinet. Tongue separated tones (left) and air-separated tones (right).

to the equilibrium position and continues to oscillate.

â&#x2014;Ś Looking into the reed bending signal in detail, a 180 phase inversion at TRC (Figure 2.15, 0.11 s) can be observed.

A possible explanation of this

behaviour could be a change of the boundary conditions of the reed. When the reed is oscillating it has only one non-free boundary condition at the point were it is mounted by the ligature. The tip of the reed remains exible and can move towards the mouthpiece.

In the moment when the tongue strokes the

reed, the boundary condition changes (Figure 2.16). Now the reed is mounted on both sides (ligature and tongue). When the returning pressure from inside the instrument acts to the reed, it deforms it in a way that the pressure can escape at the side of the reed.

Consequently, while the tongue touches the

reed, an increasing of the mouthpiece pressure deforms the reed, which might

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

â&#x2C6;&#x2019;0.6 0.5 0.0 â&#x2C6;&#x2019;1.0

â&#x2C6;&#x2019;0.5

Mouthpiece pressure

1.0

â&#x2C6;&#x2019;1.0

â&#x2C6;&#x2019;0.8

Reed bending

â&#x2C6;&#x2019;0.6 â&#x2C6;&#x2019;0.8 0.5 0.0 â&#x2C6;&#x2019;0.5 â&#x2C6;&#x2019;1.0

Mouthpiece pressure

1.0

â&#x2C6;&#x2019;1.0

Reed bending

â&#x2C6;&#x2019;0.4

0.0

0.1

0.2

0.3

0.4

0.0

0.1

Time (s)

0.2

0.3

0.4

0.5

0.6

Time (s)

Figure 2.14: Measured reed bending for tongued tones (top left) and air-separated tones (top right); Inner mouthpiece pressure for tongued tones (bottom left) and airseparated tones (bottom right) on the saxophone. explain the observed phase shift in the reed bending signal (Hofmann et al., 2012a).

Extracted parameters: Air-separated tones In the case when the player stops blowing into the mouthpiece, this loss of the energy interrupts the oscillating system. In gure 2.13 (top right) on can see a smooth fade-out of the reed bending signal. As a consequence, the inner mouthpiece pressure shows a linear decrease (Figure 2.13 middle right). If the player starts to blow again, the reed bends towards the mouthpiece lay and starts to oscillate.

2.3.4 Discussion1 This section showed that di erent articulation techniques result in di erent variation of the embouchure related parameters.

1 as

published in Hofmann et

al.

It was shown, that strain-

(2012a) and Hofmann et

al.

(2013a)

Chapter 2. Measuring musician-instrument interaction on single-reed

1.0

woodwind instruments

â&#x2C6;&#x2019;1.0 â&#x2C6;&#x2019;0.5

0.0

0.5

Reed bending Mouthpiece pressure

0.0

0.1

0.2

0.3

0.4

Time (s)

Figure 2.15: Observed phase-shift at TRC; Inner mouthpiece pressure (blue) and reed bending (red) high-pass ltered and normalized for a tongued tone transition on the saxophone.

pb n To

e gu

ng To

p ue

ng To

Figure 2.16: Blowing pressure closes the reed during sound production (top). Without tonguing, the inner mouthpiece pressure bends the reed away from the mouthpiece (middle). With tonguing the inner mouthpiece pressure deforms the reed (bottom). pb represents the blowing pressure and p the pressure in the mouthpiece.

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

gauge sensors attached to woodwind single-reeds can capture the vibrations of the reed during performance and give precise information about tongue-reed interactions without environmental interference. The obtained reed bending signal, mouthpiece pressure signal and blowing pressure signal are presented in Figure 2.13 for the clarinet (Figure 2.14 for the saxophone), focusing on a single note transition with alternatively use of tonguing or intermission of the air-stream. These measurements con rm the tonguing model of Ducasse (2003), where a tongue stroke to the vibrating reed has been described by its e ect to change the equilibrium position of the reed and its e ect to damp the oscillations. Furthermore, a di erence in the emerging transients can be observed, depending on the way that the player controls the oscillations of the instrument. From the reed bending signal it was possible to deduce that tonguing results in a stronger modulation of the reed vibrations, compared to the case of airseparated tones. Di erences also appear in the envelope of the pressure signals. Soft tonguing resulted in a smoother waveform of the pressure signals during the note transition. Observations for the e ects of tonguing on the reed behaviour were similar for saxophone and clarinet and the described characteristics occurred for all measurements obtained during these experiments. It should be noted here that the reed bending signal (depicted using the measured normalized voltages of the strain gauge sensor setup) does not directly correspond to the displacement of the reed's tip (see Section 2.2.4). Nevertheless it provides useful information on the nature of the reed oscillations and gives useful insights to articulatory player actions. The reduced complexity of the clarinet setup (simple geometry of cylindrical tube) and the enhanced measurement precision of the pressure transducers, allows a collection of physical parameters, applicable for inverse modelling techniques described in Chapter 5.

Chapter 2. Measuring musician-instrument interaction on single-reed woodwind instruments

2.4

Finger force sensors on saxophone and clarinet

This section reports preliminary test measurements with sensor reeds and with force sensors on the saxophone and the clarinet. It explains the preparation of the measurement setup used for the two empirical studies presented later in Chapter

3 and 4. Parts of this section were presented and discussed at two

international conferences and their reports were published in Hofmann et al.

(2012b, 2013c) .

2.4.1 Introduction Besides embouchure related parameters, ngerings play an important role in woodwind performance. Although the nger force to the keys does not in uence the sound directly, insights about the ngering behaviour is of interest for musicians, music teachers and also musician medicine (see Introduction Chapter 4).

2.4.2 Methods Test measurements with the saxophone2 In a rst attempt to measure nger forces to woodwind instrument keys, standard industry force sensors (Flexiforce, by Tekscan Inc.) were attached to the pearls of the rst three saxophone keys with Petro Wax (by PCB Piezotronics Inc.). This should allow to measure forces of left-hand index, middle and ring nger during performance (Figure 2.17, left). The sensing area of such a force sensor has a diameter of 9.53 mm, which covers the key-pearl and has a thickness of 0.208 mm. Simultaneous test recordings of sound, reed-signal and

am the rst author of both publications and made the following main contributions: the design of the experiments, the measurements in the laboratory, the analysis of the data, and the writing of the reports. The laboratory setup was prepared in collaboration with Alexander Mayer and Michael Weilguni. The procedure was supervised by Werner Goebl and Walter Smetana. 2 as published in Hofmann et al. (2012b) and Hofmann et al. (2013c)

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

force sensors were done to measure tongue- nger coordination in saxophone performance. A test measurement showed that when the player pressed the keys at the position of the key pearl, the sensors measured the applied force. Figure 2.18 depicts the signals of the recorded sound, the tongue action captured by the sensor reed and the nger action captured by the force sensor.

The force

curve (bottom) shows that the nger arrived at the key (FK) immediately before the tongue contacted the reed (TRC). In the following time period, while the tongue dampened the reed, the nger depressed the key until it hit the instrument body and closed the tone hole (key-bottom, KB). Then the player removed the tongue and this tongue action released the reed (TRR). The required nger force to close that key was approximately 1 N, which comes from the resistance of the key-spring. Afterwards a redundant force was applied to the key while holding it closed (5 N). An empirical investigation of nger forces on the saxophone was foreseen with this method, but failed because of interrupted data (Hofmann et al., 2012b). To further investigate the reasons of missing nger force data on the saxophone, in a second test measurement, a small and light web-camera and acceleration sensors (PCB 352C23, by Piezzotronics inc.)

were mounted to

the saxophone (Hofmann et al., 2013c). From the video recordings, it became obvious that the measured force values where in uenced by the position, the ngers arrived on the key pearl.

Some players closed the saxophone keys

with a technique, where the nger did not cover the key-pearl and thus also not trigger the force sensor. Figure 2.19 shows nger positions of individual players closing all left-hand saxophone keys to play the tone g' (Eb notation). In the photos on the left side, one can see that these players covered the key pearl with their ngertips and also triggered the attached sensor.

But the

gures on the right side show that this was not always the case. Some players only touched a part of the force sensor. From these observations, it becomes obvious that the measured force values from players not touching the sensor, are not comparable with these of players who covered the whole sensor area. Furthermore it was di cult to extract timing information of the ngers from such inconsistent force curves (Hofmann et al., 2012b). The acceleration sensors which were attached to the moving parts of the

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

keys, tracked timing information of the ngerings on the saxophone, even if the force sensor (Key Force 2) did not capture any data (see Figure 2.21).

Test measurement with the clarinet As part of the FWF project Measurement and analysis of nger forces in clarinet playing (P23248-N24), special ring-shaped force sensors for a Viennese clarinet were developed. Each sensor has three sensing elements, which allows to measure applied forces from all directions. Figure 2.22 shows the clarinet with dismounted keywork, so that the ring-shaped force sensors are visible. Details about the design and the production of the nger force sensors can be found in Weilguni et al. (2012, 2013) and in the related Ph.D. thesis Force

Sensors for the Measurement of Finger Forces in Clarinet Playing (Weilguni, 2013). A test signal, captured for a tone transition with the sensor clarinet is depicted in Figure 2.23. The signals from the three sensing elements of one sensor ring were summed (bottom).

The force curve shows that the nger

arrived at the force sensor while the tongue dampened the reed. The measured nger force was approximately 0.5 N at the time when the nger closes the tone hole.

Then the player removed the tongue from the reed (TRR) and

applied more force to the tone hole, while holding it closed (2.5 N).

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Figure 2.17: Left: Force sensors (by Flexiforce) on left-hand alto-saxophone keys; Right: alto-saxophone with mouthpiece, sensor reed and clip microphone (C419, by AKG Acoustics).

Sound

Reed bending

0.00

(FK) 0.05

●

TRR

0.10

Time (s)

0.15

0.20

TRC

0.25

TRR

0.30

applied to the key while holding it closed (5 N).

curve shows the nger-key contact (FK), the arrival of the key to the instrument body (key- bottom with 1N ) and a redundant force

woodwind instruments

key force sensor signal (bottom panel). Tongue-reed (TR) landmarks are indicated by a cross in the reed bending signal. The nger force

Figure 2.18: Recorded data for a portato tone transition on the saxophone: Audio signal (top panel), reed bending (middle panel) and

Finger Force (N)

1.0

0.5

0.0

−1.0 −0.5

−8000

−11000

−14000

TRC

Chapter 2. Measuring musician-instrument interaction on single-reed

they covered the force sensors (left) or only touched a part of it (right).

Figure 2.19: Photos of ngerpositions of four saxophone students, playing g' on an alto-saxophone. Depending on their playing technique,

36 Chapter 2. Measuring musician-instrument interaction on single-reed woodwind instruments

Chapter 2. Measuring musician-instrument interaction on single-reed woodwind instruments

Figure 2.20: Left: Player not covering the force sensor with left-hand index nger. Right: Accelerometers attached to the key-work of the saxophone to track key movements.

Figure 2.21: Sensor-reed signal (red), nger-force measurements (black) and additional tracked key-work acceleration (green), captured during performance. Acceleration sensors provide precise timing information of the ngerings.

Figure 2.22: Viennese clarinet with dismounted key-work, equipped with 6 ring-shaped force sensors (Photo by Michael Weilguni).

38 Chapter 2. Measuring musician-instrument interaction on single-reed woodwind instruments

Sound

Reed bending

2.4

2.3

2.4

TRC

TRR

2.5

Time (s)

2.6

2.7

applied to the tone hole while holding it closed.

closes the tone hole while the player damps the reed with the tongue. Similar to the observations on the saxophone, a redundant force is

woodwind instruments

summed force for sensor ring 2, measuring the left hand middle nger force (bottom panel). The nger force curve shows that the nger

Figure 2.23: Recorded data for a tongued tone transition on the clarinet. Radiated sound (top panel), Sensor reed (middle panel) and

Ring Sensor (N)

1.0

0.5

0.0

−1.0 −0.5

1.0

0.5

0.0

−1.0 −0.5

2.0

1.0

0.0

2.3

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

2.4.3 Discussion First, a measurement setup to investigate ngerings on the saxophone was presented in this section. It was shown that accelerometers at the key-work of an alto-saxophone captured the timing of the key actions more precisely than force sensors.

Additionally it turned out that standard industry force

sensors on the key pearls were not a reliable method to capture nger forces on the saxophone (Hofmann et al., 2012b).

On the saxophone, there is no

direct in uence of the ngering position to the sounding outcome, because the key-cushion closes the tone hole independent of the particular properties of the nger action. The angle and the position of the nger-key contact was highly variable for di erent saxophonists. This was con rmed by video capturing the ngers during performance (Hofmann et al., 2013c). Second, nger force signals from the sensor clarinet were inspected.

the clarinet the player has to cover the complete tone hole with the ngertip and close it properly.

Otherwise it would in uence the sound.

As a conse-

quence, the nger has to be placed on the ring-shaped force sensor developed by Weilguni (2013). An application of this sensor clarinet measurement setup will give new insights into the applied nger forces during expressive clarinet performance.

2.5

General discussion

In this chapter, di erent approaches to measure interactions between the performers and their instruments were discussed for the saxophone and the clarinet. At rst, a method to measure the reed oscillation during performance was presented. Therefore, synthetic single-reeds were equipped with strain gauge sensors. In an experiment to characterise the properties of the sensor reeds, the reed displacement and the reed bending were measured. This helped to determined that there is no in uence of the lip force to the sensor reed signal (Hofmann et al., 2013a). Subsequently, sensor reeds were used to observe the e ects of di erent articulation techniques to the vibrations of the reed (Hofmann et al., 2012a). Two di erent approaches, to capture nger actions of the player were dis-

Chapter 2. Measuring musician-instrument interaction on single-reed woodwind instruments

cussed.

First, the measurement of nger forces to the saxophone keys with

standard force sensors were examined. The recorded data was biased by individual nger-key positions of the players. To gain only timing information from the ngerings, the usage of acceleration sensors was evaluated. For studies, where nger force measurements are not required, it turned out that acceleration sensors can be a very reliable technique to investigate nger timing for future research applications (Hofmann et al., 2013c). Second, nger force measurements with the sensor clarinet (Weilguni, 2013) were investigated and a setup design to measure nger forces and tongue articulation simultaneously in clarinet performance was proposed. A combination of the developed sensor reeds, together with the nger force sensor clarinet turned out to be a suitable measurement tool for empirical studies of woodwind performances.

Chapter 2. Measuring musician-instrument interaction on single-reed

woodwind instruments

Chapter 3

Production and perception of legato, portato and staccato articulation in saxophone playing

This chapter has been published as a research article (CC-BY) in Frontiers

in Psychology (Hofmann and Goebl, 2014) . It reports and discusses two empirical studies that were carried out in the laboratory of the Institute of Music Acoustics at the University of Music and Performing Arts Vienna. The rst experiment examines di erent playing techniques on the alto-saxophone and investigates the in uence of these playing techniques on the performed timing. The second experiment is a listening experiment in which recordings of saxophone sounds with di erent articulation techniques had to be discriminated.

1 This

chapter is the full text of the journal article Hofmann and Goebl (2014), reviewed by two independent international experts in the eld (Richard Ashley, Joe Wolfe). As the rst author and the corresponding author of this publication, I made the following main contributions: the design of the production experiment, the development of the measurement methods, the technical setup of the laboratory, the recording of all participants, the development of own data processing methods for sensor reed signals, the data analysis, the script programming in R-statistics, and the writing of the research report. The listening experiment was designed in collaboration with Dominika Knapp, who also conducted the listening test and discussed the data in her unpublished diploma thesis: "Artikulation am Saxophon", Jaurovรก (2013). Werner Goebl supervised the development of the research, gave advise of statistical methods and provided guidance during the nal peer-review discussion prior to the publication of the article. 43

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

3.1

Introduction

Producing expressive sound on single-reed woodwind instruments is a highly sophisticated motor task, requiring coordination between the ngers, aural cavity and respiration (Almeida et al., 2013; Chen et al., 2011; Scavone et al., 2008). On the saxophone, a single reed of cane (or synthetic material), thinned on one end, is attached to the bottom side of a beak-shaped mouthpiece (Nederveen, 1998; Pinard et al., 2003). The player encloses the tip of the mouthpiece with his lips and blows into the tip opening. During sound production, the player's air stream excites the reed so that it oscillates related to the frequency of the impedance peak inside the instrument body (Dalmont et al., 2003; Fletcher, 1979; Almeida et al., 2010). To perform expressively on woodwind instruments, the player may use a range of parameters to shape longer sequences of tones such as onset timing and tempo or the loudness of individual tones.

An important dimension in

woodwind performance is tongue articulation, thus referring to the way the tongue controls the shape of tone onsets, tone o sets, and the connections between tones (see Bengtsson and Gabrielsson, 1983; Liebman, 1989; Krautgartner, 1982). Goolsby (1997) reported that more than

21%

of professional

band rehearsal time is spend on instructions of articulation. Articulation techniques on saxophone can be grouped in two main types: tongued articulation techniques and articulation without tonguing. Legato articulation does not involve tonguing and its sounding result is the smoothest note transition.

Herby, only changes of the ngerings determine the timing

and precision of note transitions. Professional ngering technique is required to produce smooth and clean legato tone-transitions on wind instruments (Almeida et al., 2009).

For tongued note transitions, the intensity and du-

ration of the tongue stroke to the reed de nes the sounding result.

Portato

articulation is produced by soft tongue strokes to the vibrating reed, while the player blows constantly. The sound of consecutive portato tones has been described to be close to that of legato; the subdivision of tones is very subtle.

Liebman (1989) gives instructions to the technique of tonguing on the

saxophone as follows: it is the front portion of the tongue containing muscle tissue which aps upward stroking the reed. The resulting e ect is that the

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

reed's motion and sound are momentarily stopped.

The actual sounding of

the articulation comes with the release of the reed (p.

28).

In contrast to

the soft sound of portato separated tones, staccato tones are sharp and short. These are produced by placing the tongue immediately back on the reed after the initial articulation. A consequence of these two di erent techniques of articulation (tongued, non-tongued), the timing of the performance is controlled either by tonguing or by the ngers. Timing precision in the execution of complex movement patterns is essential for musicians to produce rhythm in a sequences of tones. Palmer et al. (2009b) investigated the in uence of nger trajectories on temporal accuracy in clarinet performance, but restricted their focus on legato articulation. They reported a positive relationship between peak accelerations of nger movements and temporal accuracy of the performance, and concluded that tactile information available to the ngers supports timing control, similar to observations made with piano players (Goebl and Palmer, 2008). In woodwind performance, the ngers have to be coordinated with tongue movements to produce expressive sound.

Studies based on isochronous tapping tasks showed that a coupling

of synchronous movements operated by multiple e ectors improved temporal stability. Experiments by Ivry et al. (2002) showed that synchronous tapping with both hands improved temporal stability, compared to tapping with only one hand. Additional foot tapping enabled further temporal improvements. In the case of saxophone performance, when tongue and nger actions have to be coordinated, the multiple e ector advantage may also be the case. In line with these ndings, we hypothesise that there is a positive in uence of combined tongue- nger actions on the temporal stability in saxophone performance. In this study we will investigate how di erent articulation techniques in singlereed woodwind performance a ect performance timing. Perception and action in human motor control are strongly connected. The motor theory of speech perception argues that human understanding of speechbased auditory stimuli is based on the ability to recognize related vocal tract movements required to produce equivalent sounds (Liberman and Mattingly, 1985; Galantucci et al., 2006). Neuroimaging studies have shown that brain areas active in speech production are also active for speech listening (Fitch

et al., 1997; D'Ausilio et al., 2009; Fadiga et al., 2002). Similar observations

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

have been made for the production and perception of music (see Manto et al. (2012) for an overview). Overlap in the neural regions active when professional pianists listen to familiar pieces and the regions active when they perform these pieces has been observed (Haueisen and Knรถsche, 2001).

The link between

production and perception of music has been further discussed in the theory of auditory-motor interaction for music making (Zatorre et al., 2007). Recent research has shown that musicians are superior in judging asynchrony between sound and body movements for performances on the instruments they master than for performances on other instruments (Bishop and Goebl, 2014). Taking into account that professional saxophone players practice over a decade to acquire the skill level to produce fast tone sequences with uent articulation, we hypothesise that this motor expertise may also improve the ability to perceive articulation in saxophone performance. In this paper, we investigate articulation techniques in saxophone performances in two experiments. In a production experiment, we examine timing measures in relation to e ector combinations (tongue, nger, and both), nger movement directions (pressing for tone onsets versus releasing for tone onsets), and di erent articulation techniques (legato, portato, staccato).

a second experiment, we test whether motor expertise in a particular eld (i.e., saxophone performance) in uences the perception of di erent articulation techniques in recorded saxophone sounds.

3.2

Experiment 1: Production task

3.2.1 Methods Participants Seven female and twelve male graduate saxophone students from the University of Music and Performing Arts Vienna (N = 19, mean age = 23 years, range 18 33 years) participated in this study. On average, the participants played their instrument for 10.7 years (range = 4.5 20 years) and practised 1.9 hours per day (SD = 0.97). Eleven saxophonists reported they play classical music only, while the remaining eight participants perform usually as members of jazz ensembles.

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

Experimental design Two isochronous 24-tone melodies were designed for the experiment. melodies consisted of the same elements (Figure 3.1):

Both

The rst part (note

number 1 8) is a tone repetition, produced by only tongue actions with no change of ngerings. The following notes (9 24) require a sequential depression (melody 1) or release (melody 2) of keys by left-hand ngers. Both melodies were given as a score for alto-saxophone (sounding a major sixth lower than notated), with additional portato, staccato and legato articulation instructions. In legato articulation tone repetitions are not possible to play, thus note numbers 1 8 were omitted in the score.

Melody 1

Tongue only

Fingers + Tongue

Melody 2

Figure 3.1: Stimuli used for the production experiment. Two 24-tone melodies in E- at notation. Note numbers 1 8 require tonguing only. In melody one (top), note numbers 9 24 require sequential key-depression by left-hand ngers. In melody two (bottom) a sequential nger lifting is required to open the tone-holes of the instrument.

Equipment The experimental setup consisted of a sensor-equipped alto-saxophone, a microphone, a digital metronome, and a multi-channel recording device. Strain gauge sensors (2 mm, 120 Ohms) attached to synthetic saxophone reeds (by LĂŠgĂ¨re Reeds, Ltd.)

were used to capture the bending of the reed during

performance (Figure 3.2). The strain gauge was part of a Wheatstone quarter bridge circuit with 5 V (DC) power supply (Hofmann et al., 2013a). The sensor reed, the microphone (C414, by AKG Acoustics) and the digital metronome (KDM-1, by Korg Inc.)

were connected via BNC cables to a multichannel

analog-digital converter (DAQ LabView 2011, by National Instruments Corp.)

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

70 mm ďŹ&#x201A;at side 16.3 mm

4 mm

26 mm curved side

ďŹ&#x201A;at side

sensor area 3.1 mm

0.2 mm

curved side

Figure 3.2: Left: Synthetic alto-saxophone reed, equipped with strain gauge sensor (2 mm), glued with 4 mm distance from the tip on the at side of the reed, to avoid direct lip/tongue contact with the sensor. Right: Mouthpiece with sensor reed used in the experiments to capture reed bending during performance.

to capture the signals simultaneously. All signals were recorded onto computer hard disk (A/D conversion with sampling rate 11.025 kHz, 16 bit resolution).

Procedure The experiment was conducted in accordance with the Declaration of Helsinki: Participants gave written consent prior to the experiment, played under normal performance conditions, and received a nominal fee at the end of the experiment. In the beginning of the experiment, each player had to choose a synthetic saxophone reed out of four di erent reed-strengths (LĂŠgĂ¨re:

2.0, 2.25, 2.5,

2.75). All saxophonists were allowed to use their own mouthpiece but played on the same alto-saxophone (77-SA, by Stagg). The metronome provided the synchronisation signal on each quarter-note beat. The introduced tempi were 120 beats per minute (slow, IOI for eighth notes = IOI =

178.6 ms)

and 208 bpm (fast, IOI =

250 ms), 168 bpm (medium,

144.2 ms).

All participants got a 5

minute warm-up, to practice the melody with the metronome at a slow tempo. For the experiment, each participant played both melodies in legato, portato and staccato articulation.

They synchronised with the metronome for two

repetitions and continued playing when the metronome was muted, until the melody had been played six times in total. We recorded two trials per tempo condition, ordered from the slowest to the fastest. The experiment lasted for

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

approximately 1 hour per participant. In total 4644 tones were recorded per player (2 melodies x 2 trials x 3 tempi x 3 articulations; containing 145 tones for portato and staccato and 97 tones for legato). After their performances, the participants lled in a questionnaire about their musical background and the experiences with the sensor saxophone.

Data analysis Sensor equipped saxophone reeds were used to capture the bending of the vibrating reed during human performance. Figure 3.3 shows typical sensor reed signals in relation to the radiated sound under three di erent articulation techniques. In legato articulation (Figure 3.3a) no tongue strokes were performed, contrary to portato articulation (Figure 3.3b) and staccato articulation (Figure 3.3c). The tongue strokes to the vibrating reed are visible in the captured signals, because the tongue presses the reed towards the mouthpiece lay and thereby damps the reed vibrations. We de ne two characteristic landmarks: rst, a tongue-reed contact (TRC), when the tongue touches the reed, second, a tongue-reed release (TRR), when the tongue releases the reed and initiates the succeeding tone. The data captured during the experiment contained more than 88,000 played tones, which makes a manual transcription impossible. A multiresolution analysis (MRA) based on wavelet methods has been used successfully for the analysis of various time critical signals, ranging from medical data (i.e., ECG time series, Percival and Walden (2006)) to transcriptions of drum patterns in audio recordings (Tzanetakis et al., 2001; Kronland-Martinet et al., 1987; Paradzinets

et al., 2006). The following section discusses a landmark detection function (LDF) based on a wavelet decomposition of the sensor reed signal, where the external libraries (wmtsa, msProcess) were used in the R-statistics software package (R Core Team, 2012). The reed signal was decomposed using the Maximal Overlap Discrete Wavelet Transform (MODWT) of level

J 0 = 11.

A Daubechies least asymmetric 8-tap

lter LA(8) allows direct reference from the MODWT details to actual times in the reed signal. Figure 3.4 shows the algorithm of the LDF, working in two main steps. First, extrema in detail

e 11 (time resolution: 4t = 92.88 ms) were D

labelled. These extrema represent reed displacements caused by the tongue.

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

Hereby, maxima of

e 11 D

were labelled as TRR, because the following signal

decrease is an indicator that the player released the tongue.

Minima were

labelled as TRC because a contact with the reed must have happened before releasing the reed. Second, landmarks were shifted to the extrema in details

e 10 , with a higher time resolution (D

e9 D

and

e 8 : Ď&#x201E;8 4t = 11.61 ms). D

A spe-

cial treatment of the legato recordings was required to locate tone transitions without tongue actions.

To ensure comparable detection results, the same

MODWT analysis was applied to the legato recordings, but with an adapted LDF which worked on details

e 10 , D e9 D

and

e8 D

only. One participant's data

had to be omitted completely from the analysis, as the sensor data indicated that no tonguing was used in any of the playing conditions. To evaluate the quality of the LDF, it was tested on a small data set which contained 2020 manually annotated landmarks. Starting from the annotated ground truth, the existence and number of detected landmarks around the annotated events was checked.

A single detection was counted as one true

positive, whereas double detections were considered as one true positive and one false positive. A missing landmark was counted as one false negative. Remaining landmarks, not matched to annotated events, were counted as false positives. The standard measures precision, recall and F-measure were used.

Recall describes the completeness of the search and precision gives status about the quality of the search results. F-measure combines the two previous measures.

precision = true

positives/(true positives

recall = true positives/total number precision Âˇ recall F â&#x2C6;&#x2019; measure = 2 Âˇ precision + recall

+ false

of items

positives) (3.1)

Overall, the wavelet-based analysis gave satisfactory results of the detection tasks with F-measure

> 94%

(see A.1 for detailed values, and Hofmann et al.

2013b for further discussions of landmark detection in the sensor reed signals). To check possible in uences of the LDF to the regularity of the extracted landmarks, we calculated the time di erences of all detected landmarks to the manually annotated landmarks of the ground truth data set. A mean deviation of

0.42

ms (SD =

6.84)

showed that the detected landmarks were close to the

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

annotated landmarks.

Score-performance matching1 However, from the detected landmarks, the correct position of each performance event in the score has to be determined. Score-performance matching has been identi ed as a complex problem, especially if there are di erences between score and performance data (Gingras and McAdams, 2011; Grachten

et al., 2013). Possible reasons for these di erences can be drastic changes of expressive timing or mistakes that happened during the performance (omitted tones, wrong tones). Additionally, in our case there may be errors from the landmark detection function (false positives, true negatives) in the performance data.

Thus, it is important to verify each detected landmark, by

matching it successfully to an event in the score and to omit false detections from further analysis. A main problem in the score-performance matching with the saxophone data was, that some players omitted tones to take breath within the phrases. Additionally the F-measure of 94 % from the LDF constitutes that there may be (about 8 9) false detections in each trail with 145 tones. To identify the errors and exclude them from further analysis, a pattern-matching algorithm, based on dynamic time warping was used to match the performance data to the score. The technique of dynamic time warping allows to map events of two sequences, even though if both sequences are of di erent length. Appendix A.1 and A.2 shows two examples of the used procedure to map the performance events to the score: First, an IOI similarity matrix between performance timing and score timing was created, based on the time di erences between subsequent events (inter-onset intervals: IOI, in ms). Second, pitch information from the sound recordings of the performances were extracted with SonicVisualizer software using the Brossier (2006) pitch detection algorithm. With these pitch information, a pitch similarity matrix was created based on the pitch class distances of subsequent events from both, performance data and score information. In a third step, both matrices were combined, by multiplication. To create a nal cost matrix, the combined IOI

1 This

Ă&#x2014;

pitch similarity

subsection has been added for this thesis and is not part of the original article (Hofmann and Goebl, 2014).

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

matrix was subtracted from an all-one matrix. In a nal step, the dynamic time warping algorithm nds a path trough the nal cost matrix which has the lowest costs. A step pattern determines the rules that are allowed to the matching model. In this case, a Rabiner and Juang (1993) step pattern was chosen, which allows to omit up to three events of the sequence. With this procedure all detected events in the performances were veri ed through mapping these to events in the score.

3.2.2 Results and discussion Timing of performed melodies To examine timing of the produced sequences, we calculated inter-onset intervals (IOI, in ms), as the time interval between two subsequent TRR (onset) landmarks (IOI x

= tx+1 â&#x2C6;&#x2019; tx ).

From these IOIs we calculated the timing

error (accuracy) and the coe cient of variation (CV, precision). The timing error

(IOI obs â&#x2C6;&#x2019; IOI exp )/IOI exp

describes the relative deviation from the given

tempo. A negative value corresponds to a sequence played too fast; a positive value to a sequence played too slow. melody was calculated (CV

The temporal precision of the played

= SDIOI /M eanIOI )

to examine the regularity of

the tone-events. CV values close to zero correspond to high regularity in the sequence, while larger values indicate higher variability in the onset distribution. The average signed timing error of all performances during the synchronisation phase was close to zero

(M = 0.0077, SD = 0.032).

Figure 3.5a

(solid line) shows that all participants were able to play the melodies together with the metronome click in all three tempi.

A two-way repeated

measures analysis of variance (ANOVA) on timing error by tempo condition (metronome IOI = 250 ms, 178.6 ms, 144.2 ms) and synchronisation condition (with metronome = synchronisation, without metronome = continuation) indicated a signi cant main e ect of tempo

0.001, Îˇ = 0.662] chronisation

[F (2, 34) = 13.23, p <

as well as a signi cant interaction between tempo and syn-

[F (2, 34) = 35.77, p < 0.001, Îˇ 2 = 0.823].

Without metronome

click, participants increased the playing speed in the slow tempo condition and reduced fast tempi to a more comfortable playing speed (Figure 3.5a,

Chapter 3. Production and perception of legato, portato and staccato articulation in saxophone playing

Figure 3.3: Alto-saxophone sensor reed signals and radiated sound recorded in an anechoic chamber, showing a note transition (d2 e2 d2) under a tempo instruction of 250 ms inter-onset interval (audio sampling rate 44.1 kHz). Examples are taken from the pool of stimuli for the perception experiment in Section 3.3. a) Reed signal for legato articulation without tonguing (red) and radiated sound (black); b) Reed signal (blue) for portato articulation with tongued note onsets (tongue reed release, TRR) and note o sets (tongue reed contact, TRC) and radiated sound (black); c) Reed signal (green) for staccato articulation, with extended tongue reed contact duration and radiated sound (black).

Chapter 3. Production and perception of legato, portato and staccato

200

â&#x2C6;&#x2019;7000

â&#x2C6;&#x2019;3000

articulation in saxophone playing

200 0

D10

D7 D6

â&#x2C6;&#x2019;400 D5

â&#x2C6;&#x2019;1500 1000

â&#x2C6;&#x2019;200 100

â&#x2C6;&#x2019;400

â&#x2C6;&#x2019;300

â&#x2C6;&#x2019;400 200

â&#x2C6;&#x2019;400

D11

Reed signal

0.0

0.1

0.2

0.3

Figure 3.4: Maximal Overlap Discret Wavelet Transform of a sensor reed signal containing tongued articulation: The gure shows the input signal (top) including detected landmarks (TRC: red circle, TRR: green circle) and the details of the e 11â&#x2C6;&#x2019;5 (below). The landmark detection function labelled maxwavelet decompositon D e 11 . These positions were re ned to extrema ima (green) and minima (red) in detail D e 10 , D e 9 and D e 8. of D dashed line).

Similar observations have been reported for performances on

other instruments (e.g., piano performance, Goebl and Palmer 2013). overall temporal precision of the played sequences was high (mean CV =

0.11).

The

CV = (SDIOI /M eanIOI )

The same two-way ANOVA was calculated for

the CV and revealed a signi cant main e ect of synchronisation condition

[F (1, 17) = 19.61, p < 0.001, Îˇ 2 = 0.732] tempo and synchronisation

and a signi cant interaction between

[F (2, 34) = 6.92, p < 0.01, Îˇ 2 = 0.538].

Figure

3.5b (dashed line) shows the reduction of timing precision even for moderate playing speeds when the metronome click was removed.

Timing with multiple e ectors The melodies (Figure 3.1) were designed to consist of three distinct parts which had to be played with the ngers only (legato), the tongue only (portato note

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

0.00

â&#x2014;?

0.20 0.15

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

Slow

Medium

0.00

â&#x2C6;&#x2019;0.05

â&#x2014;?

Continuation Synchronisation

â&#x2014;?

0.10

â&#x2014;?

Timing precision

0.05

â&#x2014;?

0.05

0.10

Continuation Synchronisation

â&#x2014;?

â&#x2C6;&#x2019;0.10

Mean signed timing error

Timing accuracy (too slow)

Mean CV of timing

(too fast)

Slow

Medium

Fast

Tempo condition

Fast

Tempo condition

Figure 3.5: Timing error (a) and coe cient of variation (b), for synchronizationcontinuation playing conditions. When playing with metronome click (synchronisation phase, solid line) and without metronome click (continuation phase, dashed line). Error bars show the standard error of the mean.

repetition), and with tongue and ngers in a coordinated fashion (descending and ascending note sequence in portato articulation).

For this analysis we

restricted our data set to legato and portato recordings, because the onset detection for staccato melodies was less robust. We grouped parts of the melodies according to the e ectors required for playing and compared onset timing between these parts. A two-way repeated measures ANOVA on timing error by e ector combination and tempo, indicated a signi cant main e ect of the executing e ector

0.001, Îˇ 2 = 0.772],

a main e ect of tempo

[F (2, 34) = 25.05, p <

[F (2, 34) = 22.76, p < 0.001, Îˇ 2 =

0.757], as well a signi cant interaction between e ector and tempo [F (4, 68) = 28.39, p < 0.001, Îˇ 2 = 0.791;

Figure 3.6a]. A post-hoc pairwise t-test veri ed a

signi cant in uence of each e ector condition on the timing error (Bonferoni:

p < 0.001).

Playing with only nger actions led to faster performances than

the metronome in all three tempo conditions (mean timing error =

â&#x2C6;&#x2019;0.017).

Tongue-only actions led to slower performances compared to the metronome (mean timing error = ing conditions (M =

0.026),

0.045).

especially in the medium and the fast playUsing both e ectors in a coordinated fashion

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

(tongue + ngers) stabilised the timing error (M =

0.013),

signi cant tempo reduction in the fast tempo condition (M =

but showed a

0.058).

We ob-

served a learning e ect in the recording of the second trial for the same task

[F (1, 17) = 6.55, p < 0.05, Îˇ 2 = 0.527;

see Table 3.1].

The timing error for

combined tongue- nger actions at medium and fast tempi was signi cantly reduced in the second trial.

This indicates that professional players already

improved their tongue- nger coordination after the rst six repetitions and were able to perform the second trial with reduced timing error. A two-way repeated measures ANOVA on temporal precision (CV) by effector combination and tempo, showed a signi cant e ect of tempo

5.76, p < 0.01, Îˇ = 0.503], e ectors

[F (2, 34) =

and an interaction between tempo and the used

[F (4, 68) = 6.57, p < 0.001, Îˇ 2 = 0.528;

Figure 3.6b]. Looking at the

CV values plotted in Figure 3.6b (dotted line), we see that timing precision for the nger-only condition was lower in the slow tempo condition than in the fast tempo condition. A similar pattern appeared for the tongue- nger condition. Contrary, tones played only by tonguing (solid line) showed almost a constant irregularity over all three tempo conditions. This was con rmed by three separate one-way repeated measures ANOVAs on the CV by tempo condition. The results showed a signi cant main e ect of tempo for both conditions where ngers were involved [ ngers only: tongue + ngers:

F (2, 34) = 13.25, p < 0.001, Îˇ 2 = 0.662;

F (2, 34) = 5.51, p < 0.001, Îˇ 2 = 0.495),

but no signi cant

e ect of tempo under the condition of playing with the tongue alone.

Sep-

arate post-hoc pairwise t-tests, three for each tempo condition, showed that at the slowest and fastest tempo the tongue-only condition was signi cantly di erent from the other two conditions (Bonferoni: combined tongue- nger actions in fast tempo (p

p < 0.05),

= 0.07).

except from

There was no signif-

icant di erence between the nger-only condition and the combined tongue nger condition across all tempi. These ndings suggest that at slow tempi timing improves with combined tongue- nger actions. The observed e ect can be explained by the multiple-timer model (Ivry et al., 2002), where the timer responsible for the tongue movements is coupled to the timer of the ngers. Such a coupling of multiple e ectors has been shown to improve timing precision.

We found the opposite e ect for the fast tempo condition:

Tongue

timing deteriorated, when combined with nger movements. This fast tempo

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

0.00

â&#x2014;?

(too fast)

Slow

0.20

Fingers only Fingers + tongue Tongue only

â&#x2014;?

0.15

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

0.00

â&#x2C6;&#x2019;0.05

â&#x2014;? â&#x2014;?

Mean CV of timing

â&#x2014;?

Timing precision

0.10

â&#x2014;?

0.05

0.10

Fingers only Fingers + tongue Tongue only

â&#x2014;?

â&#x2C6;&#x2019;0.10

Mean signed timing error

Timing accuracy (too slow)

0.05

Medium

Fast

Slow

Tempo condition

Medium

Fast

Tempo condition

Figure 3.6: Timing error (a) and coe cient of variation (b), grouped by e ectors used to produce tone onsets. Error bars show the standard error of the mean.

condition (IOI = 144 ms) examined performances close to the synchronization

threshold of professional musicians (100 120 ms Repp, 2005). The measured CV values were about the same level as tones played with only ngerings. It is interesting that professional saxophonists were able to produce coordinated movements under this extreme tempo condition, but did not bene t from the coupling of the tongue to the ngers. Hence, saxophonists' tongue movements were coupled to the nger movements, even if the precision of the nger movements was worse than the precision of the tongue alone. This indicates that in saxophone playing, the timing precision of the ngers dominates the precision of the overall performance, thus overruling the timing e ects of the tongue.

Timing and the direction of nger motion To play descending tone sequences on the saxophone, keys have to be pressed, while ascending sequences require ngers to open tone holes.

To see if the

direction of nger movements (pressing down versus lifting up) in uences the timing of the performance, we contrasted (legato) sequences with a focus on key depression (Melody 1) to those focussed on lifting the keys (Melody 2). A twoway repeated measures ANOVA on timing error showed no signi cant e ect

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

Tempo (IOI)

Slow (245 ms)

Medium (178.6 ms)

Trial 1

Trial 2

Fast (144.2 ms)

Trial 1

Trial 2

Trial 1

Trial 2

-0.01010

-0.01170

-0.01402

Fingers Only

-0.03332

-0.03430

-0.00095

Fingers w. Tongue

-0.02675

-0.02760

0.01610 0.00361 0.06659 0.05029

Tongue Only

-0.00666

-0.00521

0.04549

0.03438

0.04636

0.04446

Table 3.1: Timing error for both trials of sequences performed with di erent e ectors in three tempo conditions. Bold numbers depict improved timing error (learning e ect) for combined tongue- nger actions at medium and fast playing speeds for the second trial. of the direction of nger motion nor any interactions with tempo. The same ANOVA on timing precision showed no signi cant e ects. Similar observations have been reported for clarinet performances by Palmer et al. (2009b).

Characteristics of articulation techniques We recorded reed signals of melodies with legato, portato and staccato articulation (Figure 3.3).

Whereas for legato articulation, no tongue actions

were required, portato and staccato note transitions required precise tonguing. Each of these articulation techniques allows variation within itself based on the onset and o set timing. Bengtsson and Gabrielsson (1983) discussed the resultant concepts of duration and emphasized the importance of controlling onset and o set parameters for the motion character of the rhythm. We investigated the tongue-reed contact duration (TRdur) for portato and staccato tone transitions by subtracting the TRC times (o set of previous tone) from TRR times (onset of subsequent tone, Figure 3.3). The average contact duration for portato articulation for all participants was

25.5 ms (SD = 4.1 ms,

see Figure 3.7). A one-way repeated measures ANOVA on TRdur by tempo condition showed no signi cant e ect of tempo (F (2, 34)

= 1.3, p = 0.295).

On the contrary, the same ANOVA on staccato articulation showed a highly signi cant in uence of tempo condition on the tongue-reed contact duration (F (2, 34)

= 25.2, p < 0.001).

In staccato articulation, the contact duration

(gap between tones) varies with the tempo. duration for each note transition by

We calculated the relative gap

T Rdur/IOI exp

and calculated a one-way

Chapter 3. Production and perception of legato, portato and staccato

â&#x2014;?

Portato articulation Staccato articulation

0.08

â&#x2014;?

0.06

â&#x2014;?

0.02

0.04

â&#x2014;? â&#x2014;? â&#x2014;?

â&#x2014;?

Slow

Medium

Fast

0.00

Tongueâ&#x2C6;&#x2019;reed contact duration (s)

0.10

articulation in saxophone playing

Tempo condition

Figure 3.7: Tongue-reed contact duration under di erent tempo conditions for portato articulation and staccato articulation. Error bars show the standard error of the mean.

repeated measures ANOVA on relative gap duration by tempo condition. The results showed no e ect of tempo (F (2, 34)

= 1.7, p = 0.198).

The relative

gap duration was in the range of 25 29% for all three tempo conditions (slow tempo: 0.29; medium tempo: 0.27; fast tempo: 0.25). This suggests that in portato articulation the tongue-reed contact duration remains constant, independent of the playing speed, while in staccato articulation, the relative gap duration is constant.

In uences of the measurement setup to the performances In the questionnaire, we asked how comfortable the participants felt while playing the sensor-equipped saxophone reed. The reed quality had to be rated between 1 (very good) and 7 (very bad). Results showed that the reed quality was evaluated as medium quality (M

= 3, SD = 1.5).

Participants also had to

indicate whether they felt comfortable when playing the sensor instrument or not. We tested timing accuracy and timing precision for this group e ect by separate between-subjects ANOVAs. We found no signi cant e ect between the two groups, thus, the sensor instrument did not a ect the recorded perfor-

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

mances. We also tested for e ects of self-reported handedness and skill level (years of playing the instrument), but found no signi cant e ects.

3.3

Experiment 2: Listening test

We were interested in the abilities of listeners with di erent expertise in music performance to discriminate between common articulation techniques in the sound of the saxophone. Furthermore we were interested whether motor expertise in saxophone performance would facilitate the perception of saxophone articulation.

3.3.1 Method Participants Nineteen female and twelve male (N = 31, mean age = 24 years, range = 19 32 years) students from Vienna music conservatories and Vienna universities participated in the listening study.

The group consisted of ten saxophone

players, ten musicians that play an instrument other than the saxophone and eleven non-musicians. The saxophone players had a mean of 10.5 years (range = 5 15 years, SD = 3.06) of experience in playing their instrument: eight of them also participated in the production experiment described above (Section 3.2).

The group of musicians, who did not play a wind instrument, had a

mean of 15.3 years (range = 8 22 years, SD = 3.83) of experience in musical practice of various instruments.

The group of non-musicians were students

of other elds, but 8 subjects had musical training in their early childhood, with a mean of 4.5 years. Only one of the non-musicians had experience with playing a wind instrument (the recorder).

Experimental Design In a 3

Ă&#x2014;

players

Ă&#x2014;

2 listening blocks) design, we tested which articulation techniques

Ă&#x2014;

2 (3 articulations

our participants were able to discriminate.

Ă&#x2014;

3 intervals

Ă&#x2014;

2 registers

Ă&#x2014;

We recorded note transitions of

three di erent pitch intervals (major second, minor sixth, minor sixth including register change), with legato, portato and staccato articulation, within

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

two registers by two di erent players (one of them also participated in the production experiment, the other is the rst author of this paper) on the same alto-saxophone (YAS 32, by Yamaha Corp.), using their mouthpieces (AL3, by Vandoren; Original 7*3, by Claude Lakey). Recordings were made in an anechoic chamber using a microphone (C414, by AKG Acoustics) and Labview hardware and software (DAQ LabView 2011, by National Instruments Corp.) for recording the stimuli (44.1 kHz sampling rate, 16 bit resolution). Both players used synthetic sensor equipped saxophone reeds (Section 3.2.1), to ensure tongue-reed contact in the portato and the staccato playing conditions. During the recordings, both players heard a metronome click on headphones (108 bpm for larger intervals, 120 bpm for small intervals), to produce consistent timing in the stimuli. We recorded three eighth-notes (two note transitions, see Figure 3.3) for each audio le. The beginning and the ending of the audio le was edited with a volume fade-in and fade out, to limit the sound of articulatory actions to only the two note-transitions. In total, 36 di erent audio les were comprised in the pool of stimuli. Stimuli were presented to the participants in the form of an ABX listening test on a laptop computer. A java-based software program enabled the participants to click on one of 3 buttons (A-Button, B-Button, X-Button) to play back one stimulus.

Buttons A and B contained two note transitions played

with di erent articulation techniques. Button X contained a third recording that matched the articulation used in either A or B. The question our participants had to answer was: Does X sound like A or B ? Listeners had to decide whether X was more similar to A or B. Responses and reaction time were recorded by the software.

Procedure The experimental procedure complied with the Declaration of Helsinki: Participants gave written consent prior to the experiment. All participants worked on the same laptop computer (by ASUSTeK Computer Inc.)

in a quiet en-

vironment and used the same studio headphones (K121, by AKG Acoustics). They could adjust the playback volume to a comfortable level.

Each par-

ticipant had 5 practice trials to learn how to navigate the ABX listening test software. A pop-up on the screen indicated when the actual experiment began.

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

The experiment was grouped into two blocks, each containing all 36 stimuli in four di erent listening orders. Participants were allowed to play back the stimuli as often as required to make an assured judgement.

After the rst

block was done the participants lled in a questionnaire about their musical background. Afterwards the participants made another set of judgements. The entire experiment lasted for about 30 minutes per participant.

3.3.2 Results and discussion Overall, participants from all three groups were able to accomplish the listening test with over

87%

of correct answers. A Chi-squared test revealed no signif-

icant di erence on correct answers between the two repeated listening blocks

[Ď&#x2021;2 (1) = 2.59, p = 0.11].

No e ects of listening order

or recording saxophonist

[Ď&#x2021;2 (1) = 0.05, p = 0.82]

[Ď&#x2021;2 (3) = 5.01, p = 0.17]

were found either. Due to a

labelling mistake in the playback list of stimuli, one stimulus pair had to be excluded from the results.

To convert dichotomous response data to an interval-scale level, we computed the percentage of wrong answers per participant collapsing across listening blocks and players. A two-way ANOVA on percentage of wrong answers, with articulation (type of articulation to discriminate) as within-subjects and listeners expertise as between-subjects revealed a signi cant e ect of the articulation

[F (2, 56) = 187.825, p < 0.001, Îˇ 2 = 0.933]

of the listener's expertise as a signi cant interaction

and a signi cant e ect

[F (2, 28) = 4.167, p < 0.05, Îˇ 2 = 0.479], 2

[F (4, 56) = 5.847, p < 0.001, Îˇ = 0.543;

as well see Fig-

ure 3.8a]. A two-way ANOVA on the response duration with articulation as within-subjects and listeners expertise as between-subjects factor revealed a signi cant e ect of articulation (F (2, 56)

= 67.897, p < 0.001, Îˇ 2 = 0.841),

but

no signi cant e ect of expertise or interactions (see Figure 3.8b). Focussing on articulation, post-hoc pairwise t-tests showed that results from the legato-portato listening task di ered signi cantly from the results of the other two tasks (p

< .001).

Errors occurred most often when participants

had to discriminate between legato articulation and portato articulation (25%

1 A,

sible.

B and X stimuli contained three di erent articulations: no correct answer was pos-

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

b) 20

â&#x2014;?

Musicians Nonâ&#x2C6;&#x2019;Musicians Saxophonists

â&#x2014;?

Musicians Nonâ&#x2C6;&#x2019;Musicians Saxophonists

â&#x2014;? â&#x2014;?

â&#x2014;? â&#x2014;? â&#x2014;? â&#x2014;?

legatoâ&#x2C6;&#x2019;portato

â&#x2014;?

legatoâ&#x2C6;&#x2019;staccato

portatoâ&#x2C6;&#x2019;staccato

â&#x2014;?

Response time (s)

Wrong answers (%)

â&#x2014;?

legatoâ&#x2C6;&#x2019;portato

Articulation

legatoâ&#x2C6;&#x2019;staccato

portatoâ&#x2C6;&#x2019;staccato

Articulation

Figure 3.8: Results of listening experiment: a) percentage of mistakes for discrimination task; (b) response time to accomplish the task; Participants were grouped by their expertise in music making. Error bars show the standard error of the mean. wrong answers, compared to

< 1%

wrong answers for remaining articulation

types). Additionally, Figure 3.8b shows the highest response durations for the legato-portato condition. Concering the listeners expertise, Figure 3.8 shows that non-musicians gave more wrong answers (32%) and required more time than the other two groups to respond (duration to answer per question

M = 19.35 s, SD = 19.5), followed

by non-wind-instrument players (23% wrong answers;

13.8)

and saxophonists (18% wrong answers;

M = 15.79 s, SD =

M = 11.13 s, SD = 8.7).

The

results from our listening test suggest that musical expertise alters the ability to discriminate subtle sound di erences, like between legato and portato tone transitions. The distinct sound of staccato tone transitions was well discriminated from the other two articulation techniques, by all three groups of listeners.

3.4

General discussion

This study investigated the production and the perception of articulation on the saxophone with two experiments. For the production experiment we built

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

a sensor equipped saxophone reed to monitor tongue-reed interaction in altosaxophone performance, while participants performed melodies at three tempi with di erent articulation techniques. The captured sensor-reed signals showed that for portato articulation, the tongue-reed contact duration was independent from the given tempo, whereas for staccato articulation the gap between the tones was relative to the given tempo. In legato articulation, no tongue strokes occurred and tone transitions were initiated by a change of the ngerings. Such coordination tasks occur with all wind instruments, where di erent e ectors (tongue and ngers) are required to produce one tone (e.g., ute, clarinet, trumpet). It is also the case for string instruments that the player has to coordinate di erent e ectors to produce one tone. Bowing movements with the right arm have to be coordinated with left hand ngerings. Baader

et al. (2005) looked at bow- nger coordination in violin playing and recorded tone sequences where subjects had to play a sequence of tones, in which each tone was initiated with a bow stroke and a nger change. The focus of their study was primarily on bow- nger synchronization, which was shown to be far from perfect simultaneity (50 ms), but did not lead to audible interruptions. To play larger intervals on woodwind instruments the player has to close or open multiple tone holes at the same time. This requires simultaneous nger movements, also called safe nger transitions. Almeida et al. (2009) showed that for ute performance unsafe nger transitions with approximately 21 ms already lead to audible changes in the radiated sound. Taking these studies into account, it seems that wind instrumentalists need even more precise nger movements, which additionally have to be coordinated with the movements of tongue. In our study we looked into temporal e ects of saxophone performances under di erent tempi, which were produced by di erent e ectors ( ngers only, tongue only, combined tongue- nger actions).

We found that at the slow

tempo, tone onsets produced by tongue-only actions were signi cantly less precise than tone onsets produced by ngerings only.

Highest precision was

achieved for combined tongue- nger actions. This corresponds to our hypothesis that timing precision improves for combined tongue- nger actions. However, we did not expect to see that in the fast tempo condition, tonguing alone was more precise than nger-only actions and combined tongue- nger actions

Chapter 3. Production and perception of legato, portato and staccato articulation in saxophone playing

showed a high timing variability, at approximately the same level as nger-only actions. This nding suggests that ngers play a dominant role in the overall timing of saxophone performances. In woodwind performance nger actions usually do not receive the same attention as with piano playing, where the nger movements directly produce the sound. Our observation that there is a strong in uence of nger timing on the overall timing in woodwind performance may put a new focus on further investigations of nger movements, nger trajectories, and nger forces in this domain.

With the help of sensor-equipped wind instruments and the devel-

opment of new customised sensors, useful advice for music education may be gained in future research. In the listening experiment, we observed that the articulatory sound modi cations (legato, portato, staccato) were mostly perceivable for non-musicians, musicians (not playing saxophone) and professional saxophonists.

Only the

sound of portato tongue-reed strokes was di cult to discriminate from that of non-tongued legato tone-transitions. There are two possible reasons for this. First, a brief damping of the reed vibrations does not immediately stop the standing wave in the resonator and thus only slightly modi es the radiated sound. Not all listeners notice that the reed has been stopped. Second, unsafe

nger transitions in legato playing may also cause small gaps in the sound, which non-experts may confound with portato tonguing (Almeida et al., 2009). Nevertheless, the group of professional saxophonists was superior in discriminating legato from portato sounds. An interesting observation was that the one participant, who did not use any tonguing during the production experiment and was therefore excluded from analysis there, also showed the worst results in the listening experiment (in the group of professional saxophone players).

This strengthens our assumption that expertise in the underlying

motor-actions to modify sound, facilitates perceptional discrimination of such sound modi cations. This conclusion is in line with the motor theory of speech perception (Galantucci et al., 2006): the link between perception and production of speech may also apply for the perception of articulation in saxophone music performance. As a consequence, learning to play a musical instrument enhances the ability to perceive more details of musical performance on that instrument.

Chapter 3. Production and perception of legato, portato and staccato

articulation in saxophone playing

Chapter 4

Finger forces in clarinet playing

This chapter reports a study in which nger forces of 23 skilled clarinetists were investigated during performance. The participants played eight carefully selected excerpts from the rst Weber Concerto and a technical exercise (repetition of the saxophone sound production task from Chapter 3) on a sensor equipped Viennese clarinet. The clarinet was developed at the Vienna Technical University within the FWF funded project Measurement and analysis of nger forces in clarinet playing . Special ring-shaped force sensors were developed during the project and they were attached to six tone holes on the instrument.

Details about

the construction of the low temperature co- red ceramic (LTCC) sensors can be found in the related Ph.D. thesis by Weilguni (2013). For the experiment presented in this chapter, the sensor clarinet was used for the very rst time in an empirical study. This study not only discusses the measured nger forces in clarinet playing, it also evaluates the unique nger force measurement setup.

4.1

Introduction

Clarinetists use their ngers to close or open the tone holes of the instrument during music performance. Finger actions at the tone holes are used to modulate the sounding frequency, thus quick and precise nger movements are required for professional performance (Almeida et al., 2009). Clarinetists practice for decades until they gain a professional level. Besides studying musical repertoire, practising involves technical training including scales, arpeggios and

Chapter 4. Finger forces in clarinet playing

other exercises to learn how to make well-controlled tone transitions (Mauz, 2011; Michaels, 1999). Holding a musical instrument in a proper way during playing is di cult and has to be learned. In contrast to lifting an arbitrary object (mostly with thumb and index nger, Forssberg et al., 1991), a musician has to hold the instrument in a certain position to play it. Especially with woodwind instruments, this often involves lifting the instrument weight, but remaining exible with the index- nger, middle- nger, ring- nger and pinkie nger of both hands to operate the tone holes. Holding a clarinet requires balancing the instrument on the thumbs without gripping it with the other ngers. Imbalanced stress to the thumbs has already been identi ed to be a reason for overuse syndromes with clarinetists (Diethelm, 2011). Overuse syndromes are a general problem for professional music performers across various instruments (Tubiana et al., 2005; AltenmĂźller and Jabusch, 2004). Consequently it is of interest for musicians, music teachers and medical personnel to gain insight into force related player-instrument interactions (Grosshauser and Troester, 2013; Kinoshita et al., 2012; Guillemain et al., 2010; Almeida et al., 2013; Bertsch and Mayer, 2005). With string instruments the ngertip force to the strings has been a focus of several studies. Kinoshita and Obata (2009) designed a violin with a force transducer under the ngerboard to measure string clamping force of the left hand.

The average force peak was measured with 4.5 N, and they reported

decreasing ngertip forces (1.7 N) for faster tempi.

Another aspect of their

investigations was the e ect of varying dynamics.

Although the dynamics

in violin playing are controlled by right hand bowing techniques, there was a signi cant in uence of the performed dynamics on the left-hand ngertip force. Louder playing showed increased left-hand ngertip forces. In another study, Kinoshita et al. (2012) compared expert players with novice players and found that the experts' clamping on the strings was signi cantly larger. A reason might be that holding the string tight to the ngerboard is required to produce a good sound quality and helps to avoid undesired sounds from the strings hitting the nger board. Hori et al. (2013) measured ngertip forces for holding down the strings of the guitar with about 30 50 N. An analysis of the resulting sound showed that larger force (pressing down with around 50

Chapter 4. Finger forces in clarinet playing

N) improved the tone quality and he concluded that this is the proper holding technique. However, in woodwind instruments the applied nger force to the tone hole does not in uence the resulting sound. Consequently it can be assumed that the player has to use barely minimal nger forces to close the tone holes air tight. This is also recommended by clarinet teachers (Wehle, 2007). Investigations of ancillary gestures in expressive clarinet performance have shown that a lot of body movements happened, that were not related to economic control over the instrument only (Desmet et al., 2012; Caramiaux et al., 2012). Circular movements of the clarinet bell, head and shoulder movements of the performer, bending at the waist or with the knees, apping the arms and stepping with the feet were typical ancillary gestures. Wanderley et al. (2005) also observed that for each performer similar movements appeared for repetitions of the same piece. Palmer et al. (2009a) explained further that ancillary gestures in clarinet playing were related to acoustic features of musical expression. Such an embodiment to the musical expression might also a ect the nger force pro les during expressive music performance. Furthermore, to play with loud dynamics or in the high register, the player has to blow with more pressure into the instrument.

Does such a tension of the respiratory

system in uence the nger forces applied to the tone holes? Motion capture studies focusing on the ngerings of clarinetists reported that during di cult and fast sequences, players used larger nger movements and increased accelerations at nger-key contact (Palmer et al., 2007, 2009b). Palmer et al. (2009b) suggest that there is an e ect of tactile feedback on temporal accuracy. Increased sensory information to the performers ngers seemed to improve the timing accuracy of the played sequence. Similar observations for ute performance were made and may support this hypothesis (Almeida

et al., 2009). Hofmann and Goebl (2014, see Chapter 3) measured the e ect of di erent e ectors used (tongue-only actions, nger-only actions, combined tongue- nger actions) to play melodies on the alto-saxophone. The observations made have partly supported the multiple-timer model by Ivry et al. (2002). In the case of the slow tempo condition, combined tongue- nger actions stabilized the overall timing. For fast tempi, nger actions showed a dominant in uence on

Chapter 4. Finger forces in clarinet playing

the performance timing. Furthermore, in that experiment two melodies were used as stimuli. One melody focused on pressing the keys of the instrument and the other focussed on releasing the keys. No in uence of direction of nger movements on the performed timing was found, similar to reports from other studies Palmer et al. (2009b). Consequently, a repetition of this experiment, does not require to measure both directions of nger movements. However, a repetition of this study on a di erent single-reed instrument, would allow to verify the observations made with saxophonists. Moreover, it would be interesting to proof if a clarinetists, with a classical music background show similar results as the group of jazz and classical saxophonists.

With

a nger force sensor clarinet it will additionally be possible to measure the applied force to the tone holes and report nger forces used with this technical exercise, requiring actions close to the synchronization threshold of professional musicians (Repp, 2005). The foreseen study focusses on measuring the nger forces of index nger, middle nger, and ring nger of both hands during expressive clarinet performance and will also include a repetition of the technical exercise from the saxophone study. A special Viennese Clarinet, equipped with six force sensor rings at the tone holes of the instrument (Weilguni, 2013) together with strain gauge sensor equipped single-reeds (Hofmann et al., 2013a) will be used to capture nger force pro les and tongue actions of professional clarinetists.

4.2

Performance experiment

4.2.1 Methods Participants Ten female and thirteen male clarinetists (N = 23, mean age = 26.9 years, range 19 45 years) participated in this study.

On average the participants

played the clarinet for 17.26 years (range 9 37 years) and reported to practice with their instrument for about 2.78 hours per day (SD = 1.16). Seventeen participants were students from the University of Music and Performing Arts Vienna and six participants were professional performers, ve of them were additionally teaching at an academic institution.

Chapter 4. Finger forces in clarinet playing

Experimental design For the experiment three di erent performance tasks were prepared: a warmup task, an expressive performance task, and a technical exercise task. For the warm-up task, a simple melody was designed with a focus on sequentially opening and closing the six sensor equipped tone holes ( nd score in Appendix B.1). For the expressive performance task, excerpts from concert clarinet literature were chosen to measure nger forces under di erent musical situations. In a 2

Ă&#x2014; 2 Ă&#x2014; 2 Ă&#x2014; 2 design (register:

low high; tempo: slow fast, dynamics: soft

loud, expression level: low high), nger forces during clarinet performances will be investigated.

Eight excerpts from the Clarinet Concerto No.

1 in F

minor (Op. 73) for clarinet in Bb, by C.M.v.Weber were selected ( nd score in Appendix B.2 and B.3). This particular piece was chosen, because it belongs to the standard repertoire of clarinet students at the University of Music and Performing Arts Vienna.

Each selection ful lled one of three testing condi-

tions (see Appendix B.1), the fourth testing condition will be introduced by the experimenter during the experiment. For the technical exercise task, an isochronous 23-tone melody was designed for the experiment (Appendix B.4), similar to the melody used in the saxophone study (see Chapter 3 or Hofmann and Goebl, 2014): The rst eight notes of the melody are a repetition of the same tone. The notes after (9 23) are performed by closing the tone holes with the ngers of the left hand. The score was prepared for clarinet in Bb and included the articulation instructions for portato, staccato and legato playing. Because of the reason that note repetitions in legato are not possible to play, the notes 1 8 had to be removed from the score for this playing condition.

Equipment The following setup was chosen for this experiment: a sensor-equipped Viennese clarinet in Bb (see Figure 4.2), a microphone, a digital metronome, a small camera, and a multi-channel recording device. To measure the reed vibrations and the tongue actions of the players during performance, sensor reeds were prepared for this study (see Chapter 2.2 for details). Six ring-shaped force sen-

Chapter 4. Finger forces in clarinet playing

Figure 4.1: Ring-shaped sensor attached to a tone-hole of the clarinet: (left) without polymer ring; (right) with polymer ring glued on three soda lime glass balls with epoxy adhesive. The surrounded key-work has been dismounted for this picture (Photos by Michael Weilguni).

Figure 4.2: Viennese Sensor Clarinet in Bb (Model38, by Foag), with six force sensor rings mounted to the tone holes and a sensor clarinet reed.

sors were mounted around the six front tone-holes of the Bb clarinet (Model 38, by Martin Foag, Hafenhofen), to measure the nger forces of index- nger, middle- nger, and ring- nger of both hands during clarinet performance (Weilguni, 2013). The sensor reed, the force sensor rings, the microphone (d:vote 4099, by DPA Microphones) and the digital metronome (KDM-1, by Korg Inc.)

were connected via BNC cables to a multichannel analog-digital con-

verter (DAQ LabView 2011, by National Instruments Corp.)

to record the

signals simultaneously (A/D conversion with sampling rate 11.025 kHz, 16 bit resolution). The data was captured onto computer hard disk .

Chapter 4. Finger forces in clarinet playing

Procedure The experiment was conducted in accordance with the Declaration of Helsinki. All participants gave their informed consent to the procedures of the study and received a nominal fee after the experiment. In the beginning of the experiment each participant chose a sensor reed according to the preferred reed strength (LĂŠgĂ¨re Nick, Bb clarinet, german cut, reed strength: 2.5 4.5). Participants were allowed to use their own mouthpiece but played on the same sensor equipped clarinet in Bb. Before the recordings started, participants had ve minutes time to adjust the setup according to their needs. The main experiment contained ve tasks:

First, the warm-up task was

performed to a metronome signal with 120 beats per minute on each quarternote beat Appendix B.1. Second, the expressive performance task, containing eight excerpts from the Weber Clarinet Concerto, was played.

We recorded

four trials for each excerpt. The instructions for the rst two trials were, to focus on technical aspects of playing (e.g., correct tones, synchronisation to the provided metronome signal). For the remaining two trials, the metronome signal was muted and the instructions were to play expressively (choose own tempo, focus on dynamics), like in a concert situation.

Third, participants

lled in a questionnaire about their musical background and gave feedback about playing the sensor equipped clarinet. For a technical exercise (fourth) task, each participant played the 23-tone melody (Appendix B.4) in legato, portato and staccato articulation in three tempo conditions (slow, IOI for eighth notes = fast, IOI =

144.2 ms).

a quarter-note level.

250 ms; medium, IOI = 178.6 ms;

The metronome provided the synchronisation signal on For the rst two repetitions, participants synchronised

with the metronome. When the metronome was muted, they continued playing, until the melody was played six times in total. We recorded two trials per tempo condition, which were ordered always from the slowest to the fastest (to repeat the procedure from Hofmann and Goebl, 2014).

In a nal, fth

step, the warm-up task was performed again to the metronome signal. The experiment lasted for approximately 1 hour per participant.

Chapter 4. Finger forces in clarinet playing

Data analysis Each of the six force sensor rings attached to the tone holes of the sensor clarinet, contained three measurement cells with

120â&#x2014;Ś

distance (see Figure 4.1).

Calibration values for the measurement cells were provided by the manufacturers from the TU Vienna (Weilguni, 2013), to convert the measured output voltage to force in Newton (N). During the experiment it turned out that the sensors were fragile. For example sensor 4 (right-hand index nger) broke completely through the mechanical burdening while recording the sixth participant. All three measurement cells from this sensor fell o . The measurement cells were glued with conductive silver/epoxy adhesive, into the sensor ring. Even putting them back in changed the behaviour of the sensor. Figure 4.3 (bottom) shows that the force values captured by Sensor 4 (dotted line), are comparable with the measurements from the other sensors until player 6, and than for player 7 9 it produces wrong data. For player 10 13 and 17 23 the sensor did not capture any data at all. Consequently all measurements from sensor 4 were removed from further analysis steps. Unfortunately, small artefacts occurred in the data of all sensors at some point. A reason for this were the cable connections between the upper and the lower part of the clarinet, going over the right hand thumb of the player.

If the player touched the cables, this inferred the data.

Fur-

thermore, the sensors showed the tendency to drift during the study. The 0 N had to be calibrated for each recording session individually, and in some cases, they also drifted during one experiment. As a consequence, all the captured data of each measurement cell had to be carefully inspected. If artefacts or such a drift occurred, the measurement cell was excluded from the analysis. The values of each working measurement cell from one sensor ring were summed to calculate the force applied by each nger to the sensor ring. Examples for nger force pro les are shown in Appendix B.6 and B.7. If data from one or two measurement cells was excluded, weighted sums were used to compensate this.

When all three measurement cells produced

wrong data, no nger force pro le for this sensor ring exists. Figure 4.3 shows, that at the beginning of the experiment (ID 1) all 6 sensors worked properly, and for the last participant (ID 23) only 3 sensors produced reliable data.

Chapter 4. Finger forces in clarinet playing

From the nger force pro les, the regions where the player closed the tone hole were selected automatically by using a threshold. For the selected regions, the peak force (Fmax ) and the average force (Fmean ) were calculated and the required force to press down the key-work around the tone hole was added. Post processed nger force measurement results were saved per nger and per recorded trial for further statistical analysis. To examine timing properties of the technical exercise task, tongue-reed contact (TRC) and tongue-reed release (TRR) landmarks were extracted from the captured reed signals.

The landmark detection function and the score-

performance matching were similar to the method reported for the saxophone study in Chapter 3.2.1 (p. 49).

Mean Finger Force (N)

●

● ●

●

● ●

●

● ●

●

● ●

●

● ●

●

● ●

●

Participant ID

●

● ●

●

Participant ID

●

● ●

Peak Finger Forces per Participant

● ●

●

Mean Finger Forces per Participant

●

● ●

●

● ●

●

● ●

●

1 2 3 4 5 6

●

1 2 3 4 5 6

Ring Sensors

●

Ring Sensors

Bottom: Average peak forces (Fmax in N) measured for each participant.

index nger, 2 = l-h. middle nger, 3 = l-h. ring nger, 4 = right-hand index nger, 5 = r-h. middle nger, 6 = r-h. ring nger).

Clarinet Concerto No. 1 by Weber). Each line represents the force data captured for one nger during the entire task (1 = left-hand

Figure 4.3: Top: Average nger forces (Fmean in N) measured for each participant during the expressive performance task (excerpts from

Peak Finger Force (N)

76 Chapter 4. Finger forces in clarinet playing

Chapter 4. Finger forces in clarinet playing

4.2.2 Results Finger forces during expressive performance task The average nger force measured for all participants during the expressive performance task (excerpts from Weber Concerto No. Newton (N) with a standard deviation of

SD = 0.37.

1) was

Fmean

= 1.17

Measured peak nger

forces (Fmax ) over all recorded trails varied between 0.84 N and 12.95 N with an average of 3.05 N and a standard deviation of 2.0 N. Figure 4.3 (top) shows the average measured nger force for each participant's nger operating on a sensor equipped tone hole. Each line represents the captured data of one force sensor ring during the experiment. Looking into the average peak nger forces, it is obvious that participant 4 showed the highest peak nger forces captured during the expressive performance task (above 12 N). To compare individual nger force pro les, Appendix B.6 (ID 2) and Appendix B.7 (ID 4) depict force measurements for excerpt A

of the expressive performance task . Note in Appendix B.5 (right) that participant 4's ngertips already changed its color when closing the tone holes. A one-way analysis of variance (ANOVA) on

Fmean

with participants' sex

as between-subjects factor revealed no signi cant e ect of the player's sex

[F(1,21) = 0.703, p = 0.411], although females average nger forces were slightly lower than the nger forces of the male participants (female male

Fmean

= 1.1 N;

= 1.2 N). Taking the professional level of the players (students vs.

professionals) into account, showed also no signi cant e ect with an one-way ANOVA on factor

Fmean

testing participants' professional level as between-subjects

[F(1,21) = 1.081, p = 0.310].

To examine nger forces under di erent situations of musical expression, participants played eight excerpts with di erent musical instructions given by the score (register: low high; tempo: slow fast, dynamics: soft loud). Each sequence was performed under two di erent playing instructions from the experimenter. In one condition, the clarinetists had to focus on technical aspects of playing (e.g., correct pitch and precise note length, synchronisation to the

1 Raw

data of left-hand index nger force pro les, captured for all participants performing both trials of excerpt A in the highly expressive playing condition are plotted in Appendix B.8, B.9, and B.10.

Chapter 4. Finger forces in clarinet playing

1.3 1.2 1.1

â&#x2014;?

0.9

1.0

Mean Finger Force (N)

1.4

1.5

Expression

Low

High Expression Level

Figure 4.4: Mean nger forces (Fmean in N) for expressive performance task, increased with a higher level of musical expression. Error bars show the standard error of the mean.

metronome), where in the second condition the clarinetists were instructed to play with a high level of musical expression, similar to a concert performance in front of an audience (expression level: low high). A one-way repeated measures ANOVA on

Fmean

by expression level indi-

cated a signi cant main e ect of expression level during performance

26.2, p < 0.001, Îˇ 2 = 0.74].

[F(1,22) =

Figure 4.4 shows that force values increased for

performances with a higher level of expression. To examine nger forces how they would appear in performance situations, only the trials recorded in the high expression level performance condition are further analysed and discussed. A three-way repeated measures ANOVA on

Fmean

by register, dynamics, and tempo revealed three signi cant main e ects

(register

[F(1,22) = 19.04, p < 0.001, Îˇ 2 = 0.681];

0.05, Îˇ 2 = 0.414];

tempo

[F(1,22) = 4.56, p <

[F(1,22) = 5.22, p < 0.05, Îˇ 2 = 0.438;

and two interactions (register and tempo and tempo

dynamics

[F(1,22) = 23.52, p < 0.001;

see Figure 4.5])

[F(1,22) = 18.39, p < 0.001]; dynamics

see Figure 4.6]).

Playing in the high

register led to increased nger forces, as well as louder dynamics (Figure 4.5 a, b). Overall, sequences that required playing at fast tempi showed reduced nger force values compared to the slower parts (Figure 4.5c).

Taking the

Chapter 4. Finger forces in clarinet playing

1.5 1.4 1.3 1.1

1.2

â&#x2014;? â&#x2014;?

0.9

1.0

1.1

â&#x2014;?

Mean Finger Force (N)

1.4 1.3 1.2

â&#x2014;?

1.0

Mean Finger Force (N)

Dynamics

1.5

Low

High

Soft

Loud

Dynamic condition

Tempo

1.3 1.2

â&#x2014;?

1.1

â&#x2014;?

0.9

1.0

Mean Finger Force (N)

1.4

1.5

Slow

Fast

Tempo condition

Figure 4.5: Mean nger forces (Fmean ) for di erent musical instructions. a) Playing in high register led to increased mean nger forces. b) Increasing dynamics increased nger forces. c) Playing in a fast tempo resulted in reduced nger forces. Error bars show the standard error of the mean.

Chapter 4. Finger forces in clarinet playing

1.8 Slow Tempo Fast tempo

●

1.6

Slow Tempo Fast tempo

●

● ●

●

1.0

●

1.4

●

1.2

1.4

1.6

●

Mean Finger Force (N)

●

Mean Finger Force (N)

Dynamics and Tempo

1.8

Low

High

Soft

Loud

Dynamic condition

Figure 4.6: Interactions between register, dynamics and tempo of performances. a) In slow tempo, participants' nger forces increased signi cantly when playing in high register but were nearly constant for fast tempi (solid line). b) Loud dynamics led to increased nger forces in slow tempo, but showed the opposite behaviour in fast tempo. Error bars show the standard error of the mean. signi cant interactions into account, separate post-hoc t -tests, two for each tempo condition were necessary. The result indicates that in the slow tempo condition, register (p

< 0.001)

and dynamics (p

< 0.001)

had a signi cant

in uence on nger force in terms that increased dynamics and playing in high register showed increasing nger forces (Figure 4.6a, b; dashed line). Whereas in the fast tempo condition only dynamics (p

< 0.05)

signi cantly in uenced

the nger forces, and register had no further in uence (p

= 0.69;

Figure 4.6a,

solid black line). It is interesting to note in Figure 4.6b that depending on the tempo the direction of in uence through dynamics changed.

In slow tempo

(dashed line), loud dynamics led to increased nger forces, whereas in fast tempo (solid line), nger forces decreased with louder dynamics.

Self-evaluation of nger forces In the questionnaire, participants were asked to self-evaluate their nger forces, without knowing any results of the measurement. Participants had to report on a seven-step rating scale (from 3 to +3), whether they think their own nger

2.0

â&#x2014;?

1.5

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;? â&#x2014;? â&#x2014;?

1.0

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;? â&#x2014;? â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

0.5

â&#x2014;?

0.0

Measured Mean Finger Force (N)

2.5

Chapter 4. Finger forces in clarinet playing

â&#x2C6;&#x2019;2

â&#x2C6;&#x2019;1

Self Reported Finger Force

Figure 4.7: Self-estimation of nger forces (from 3 to +3) compared to measured Fmean nger forces. force is below average (from 3), on average (0), or on a higher force level (up to +3). We correlated the measured nger forces (Fmean ) for expressive performance with the answers of the questionnaire. The measured nger force showed a high correlation with the players' self-evaluation of their nger force

[r = .64, p < 0.001;

Figure 4.7].

We also asked the participants to report any physical discomfort in relation to clarinet playing. Non of the participants reported permanent physical discomfort.

For the participants who reported only occasionally discomfort

during playing (19%) or during and after playing (24%), an one-way repeated measures ANOVA on

Fmean with discomfort as between-subjects factor did not

reach a level of signi cance

[F (1, 19) = 3.62, p = 0.07, Îˇ = 0.4],

but showed a

medium e ect. However, participants who reported physical discomfort, selfevaluated their nger forces as above average (+1.2) and also showed larger nger force values during the experiment (Fmean reporting no discomfort at all (Fmean

= 1.4N )

compared to those

= 1.08N ).

When asking the participants if they would be interested in using a force sensor clarinet for practising or for teaching, the results showed no clear trend (self-use: yes = 43 %; teaching: yes = 48 %). However, looking at the two groups of players separately (reported physical discomfort or not) showed that

Chapter 4. Finger forces in clarinet playing

the group of players with problems had more interest (70%) in using such a measurement device than the other group (27%).

Timing of the performed melody in the technical exercise task1 To examine the timing in the technical exercise task, the inter-onset intervals between subsequent TRR (oneset) landmarks in the reed signal were calculated (IOI x

= tx+1 â&#x2C6;&#x2019; tx ).

From these IOIs, the timing error ((IOI obs

â&#x2C6;&#x2019;

IOI exp )/IOI exp ) and the coe cient of variation (CV = SDIOI /M eanIOI ) was derived (similar to the analysis in Hofmann and Goebl, 2014).

The timing

error (timing accuracy) is a measure for the deviation from the given tempo, where negative values indicate a sequence played too fast and positive values correspond to a sequence played too slow. The coe cient of variation (CV, timing precision) depicts the regularity of the tone onsets. Figure 4.8a shows the timing error for all performed melodies, grouped by synchronisation condition for all three tempi. The solid black line shows that

all clarinetists were able to perform the melodies to the metronome click . Muting the metronome click resulted in participants playing too fast in the slow tempo condition and too slow in the fast tempo condition (dashed line). A three-way repeated measures ANOVA on timing error by tempo condition (slow, medium, fast) and synchronisation condition (with metronome click = synchronisation, without metronome click = continuation) as within-subjects and participants' professional level (students vs.

professionals) as between-

subjects factor con rmed a signi cant main e ect of tempo

p < 0.001, Îˇ 2 = 0.695] synchronisation

[F (2, 40) = 18.645,

as well as a signi cant interaction between tempo and

[F (2, 40) = 17.093, p < 0.001, Îˇ 2 = 0.679]

professional level

[F (2, 40) = 6.232, p < 0.01, Îˇ = 0.487].

professional level on timing accuracy (p

= 0.424)

and tempo and

No main e ect of

nor any other signi cant

interactions were found. The performed sequences showed a high timing precision (mean CV = 0.09). The same three-way ANOVA calculated for the CV indicated only a signi -

1 In

this subsection, the same analysis methods as with the saxophone study (Hofmann and Goebl, 2014) are used, in order to gain comparable data. 2 Data of one participant had to be omitted from the analysis, because the sensor data was interfered by artefacts.

Chapter 4. Finger forces in clarinet playing

â&#x2014;?

(too fast)

Slow

0.20 0.15

â&#x2014;?

â&#x2014;? â&#x2014;? â&#x2014;?

â&#x2014;?

Slow

Medium

â&#x2014;? â&#x2014;?

0.00

â&#x2C6;&#x2019;0.05

â&#x2014;?

Continuation Synchronisation

â&#x2014;?

0.10

0.00

â&#x2014;? â&#x2014;?

Timing precision

0.05

â&#x2014;?

0.05

0.10

Continuation Synchronisation

â&#x2014;?

â&#x2C6;&#x2019;0.10

Mean signed timing error

Timing accuracy (too slow)

Mean CV of timing

Medium

Fast

Tempo condition

Fast

Tempo condition

Figure 4.8: Timing error (a) and coe cient of variation (b), for synchronizationcontinuation playing conditions for the technical exercise.

When playing with

metronome click (synchronisation phase, solid line) and without metronome click (continuation phase, dashed line). Error bars show the standard error of the mean.

cant main e ect of tempo

[F (2, 40) = 3.611, p < 0.5, Îˇ 2 = 0.391].

Figure 4.8

shows that the regularity of subsequent events deteriorated with increasing tempo.

In comparison to the measurements on the saxophone (see Chapter

3.2.2), synchronisation condition did not reach a signi cant level (p

= 0.0584),

because the clarinetists' regularity of events was higher than with the saxophonists, especially in the continuation condition. The timing quality (timing error, timing precision) for both participant groups (students and professionals) was in general the same.

A possible reason that no signi cant e ect in

timing quality between the two groups (students vs. professionals) was found, might be that the group of professional players was small (N = 6) and the participating students played their instrument on average for 13 years.

All

students had successfully passed the music university entrance exams which indicates that they already perform on a high level.

Chapter 4. Finger forces in clarinet playing

Timing with multiple e ectors in the technical exercise task1 The melody of the technical exercise consisted of three parts. The rst part required only nger actions (legato), the second part required only tongue articulation (portato tone repetition) and the remaining notes required si-

multaneous tongue and nger actions . Onset timing of the three parts was compared by analysing timing error and timing precision (Figure 4.9).

two-way repeated measures ANOVA on the timing error by e ector combination and tempo, showed a signi cant main e ect of the executing e ector

[F (2, 42) = 8.95, p < 0.001, Îˇ 2 = 0.547],

as well as a main e ect of tempo

[F (2, 42) = 18.52, p < 0.001, Îˇ = 0.685] (no interactions).

A post-hoc pairwise

t-test showed that timing with only nger actions was signi cantly di erent from only tongue actions (Bonferoni: actions (Bonferoni:

p < 0.001).

p < 0.001)

and combined tongue- nger

Similarities to the results of the saxophone

experiment (see Chapter 3.2.2 or Hofmann and Goebl, 2014) were found in the clarinet data: Playing with the ngers alone led to faster performances and playing with the tongue alone led to a reduction of the tempo compared to the metronome. Combining both e ectors stabilized the timing error. A two-way repeated measures ANOVA on temporal precision (CV) by e ector combination and tempo, resulted in two signi cant main e ects but showed no interactions. Figure 4.9b depicts how timing precision deteriorated with increasing tempo

[F (2, 42) = 20.18, p < 0.001, Îˇ 2 = 0.7]

by the combination of e ectors

and was also a ected

[F (2, 42) = 6.317, p < 0.01, Îˇ 2 = 0.481].

The

measured values for nger-only timing precision were on approximately the same level as with the saxophone study and showed a similar tempo pattern (clarinetists' mean CV = 0.097, Figure 4.9b; saxophonists' mean CV = 0.10, Figure 3.6b). A main di erence in the results of both studies was that the clarinetists' tongue-only actions showed a much higher temporal precision, especially in the slow and the medium tempo condition (clarinetists' mean CV = 0.074, saxophonists' mean CV = 0.11). Consequently, nger-only actions appeared to have the lowest timing precision compared to the other e ector combina-

1 In

this subsection, the same analysis methods as with the saxophone study (Hofmann and Goebl, 2014) are used, in order to gain comparable results. 2 For this analysis the data set was reduced to legato and portato performances only.

Chapter 4. Finger forces in clarinet playing

â&#x2014;?

0.20

(too fast)

Slow

0.15

â&#x2014;?

â&#x2014;? â&#x2014;? â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

0.00

â&#x2C6;&#x2019;0.05

â&#x2014;?

Fingers only Fingers + tongue Tongue only

â&#x2014;?

0.10

0.00

â&#x2014;? â&#x2014;?

Timing precision

0.05

â&#x2014;?

0.05

0.10

Fingers only Fingers + tongue Tongue only

â&#x2014;?

â&#x2C6;&#x2019;0.10

Mean signed timing error

Timing accuracy (too slow)

Mean CV of timing

Medium

Fast

Slow

Tempo condition

Medium

Fast

Tempo condition

Figure 4.9: Timing error (a) and coe cient of variation (b), grouped by e ectors used to produce tone onsets. Error bars show the standard error of the mean.

tions across all tested tempo conditions (Figure 4.9b, dotted line), although it showed the same CV than with the saxophonists. However, similar to the observations on the saxophone was that combined tongue- nger actions improved timing in the slow tempo condition.

Furthermore, at fast tempo, combined

tongue- nger actions were closer to the precision of the ngers-only condition. Separate post-hoc pairwise t-test, three for each tempo condition, con rmed that at slow tempo, combined tongue- nger actions were signi cantly di erent from nger-only actions (Bonferoni:

p < 0.01).

This observation would sup-

port the multiple timer theory (Ivry et al., 2002), where a coupling of multiple e ectors leads to improved timing. However, in the medium and the fast tempo condition there was no signi cant di erence between combined tongue- nger actions and nger-only actions (Bonferoni:

p = 1.0) observable.

Moreover, the

timing of the ngers overruled the timing of the tongue when both e ectors were used in a coordinated fashion. In line with the observations on the saxophone, this indicates that nger movements play a dominant role in the overall timing precision of clarinet performance.

Chapter 4. Finger forces in clarinet playing

Timing and nger forces in the technical exercise task Looking at the measured left-hand nger force for the technical exercise task, participants used lighter ngering on the sensors (Fmean = 0.64 N) compared

to the expressive performance task . A two-way ANOVA on

Fmean

condition and professional level showed a main e ect of tempo

16.751, p < 0.001, Îˇ 2 = 0.675]. a level of signi cance

[F (2, 42) =

Professional level of the players did not reach

[F (1, 20) = 1.653, p = 0.213]

on the used nger force (Îˇ

by tempo

= 0.276).

but showed a small e ect

Figure 4.10 depicts that the nger

forces of professional players were softer compared to those of the students, Separate post-hoc t-tests

especially in the slow and in the medium tempo.

(Bonferoni) for each tempo condition con rmed this observation for the slow and the medium tempo (slow tempo: fast tempo:

p = 0.153).

p < 0.05;

medium tempo:

p < 0.05;

In comparison to the expressive performance task,

faster tempi for the technical exercise led to increased nger forces. A possible reason might be that the technical exercise melody was be played at only one (comfortable) dynamic level, without musical expression and in the low register of the instrument. Weilguni (2013) hypothesed in his Ph.D. thesis that too intense pressing of the clarinet keys may result in unwanted uctuations in tone durations and limitations in the maximum playing tempo.

To test whether there is a

correlation between the applied nger force to the keys and the maximum playing tempo, we assume that higher nger force must result in a positive mean signed timing error of performances.

For each tempo condition, the

mean timing error was correlated with the measured nger forces (Fmean ). No signi cant correlation was found in our data (slow tempo:

0.499;

medium tempo:

r = 0.08, p = 0.590;

fast tempo:

r = â&#x2C6;&#x2019;0.10, p =

r = 0.14, p = 0.369).

The same correlation for the coe cient of variation was calculated to test the hypothesis of unwanted uctuations in tone durations, by correlating timing precision with nger forces (Fmean ). No such correlation was found in our data (slow tempo: tempo:

1 To

r = â&#x2C6;&#x2019;0.04, p = 0.843;

r = â&#x2C6;&#x2019;0.14, p = 0.548).

medium tempo:

r = 0.09, p = 0.675;

fast

From a comparison with the measurements of

put a focus on the ngering technique the dataset was restricted to the performances with the nger-only condition.

Chapter 4. Finger forces in clarinet playing

1.0

Finger forces in technical excercise Students Professionals

●

Mean Finger Force (N)

0.8

●

0.6

● ●

●

● ●

0.0

0.2

0.4

●

Slow

Medium

Fast

Tempo condition

Figure 4.10: Average left-hand nger forces for technical exercise task, grouped by the professional level of the participants. Professional players used less nger forces than clarinet students. the expressive performance task (Fmean = 1.2 N for performances with a high expression level), it is obvious that the participants used less nger forces in the technical exercise task already (Fmean = 0.64 N). Figure 4.11 depicts the force measurements (top) for each participant. The bottom panel allows to compare the timing precision between all participants and it is visible that participant 14's timing precision was superior in comparison to the others. From the questionnaire it was possible to determine that

this was the only participant who reported left-handedness . It is interesting to see that the timing precision of the left-handed player was superior (CV = 0.053) in comparison to the average precision of all participants (CV mean = 0.098).

The reason why this technical exercise focussed on left-hand ngerings was through a limitation in the number of sensors available during the pre-saxophone study. To collect comparable data, the melody was not changed for the successive experiments. Besides, sequences with left-hand nger actions exist in musical scores as well and have to be practised and performed by every player.

Mean Finger Force (N)

Mean signed timing error

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

IDs

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

IDs

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

IDs

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

Timing precision per participant in technical excercise task

â&#x2014;?

Timing accuracy per participant in technical excercise task

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

Mean finger forces per participant in technical excercise task

â&#x2014;?

â&#x2014;? â&#x2014;?

â&#x2014;?

â&#x2014;? â&#x2014;?

Slow Medium Fast

â&#x2014;?

Slow Medium Fast

Tempo condition

â&#x2014;?

Tempo condition

Slow Medium Fast

Tempo condition â&#x2014;?

â&#x2014;?

coe cient of variation (bottom panel) for the same performances.

under the legato articulation condition (Bar 2 and 3 of score in Figure B.4) in three di erent tempi. Timing error (middle panel) and

Figure 4.11: Top: Average left-hand nger forces (Fmean ) measured for each participant during the technical exercise task, performed

Mean CV of timing

2.5

2.0

1.5

1.0

0.5

0.0

0.2

0.1

0.0

â&#x2C6;&#x2019;0.1

â&#x2C6;&#x2019;0.2

0.5

0.4

0.3

0.2

0.1

0.0

88 Chapter 4. Finger forces in clarinet playing

Chapter 4. Finger forces in clarinet playing

E ect of fatigue over the duration of the experiment (warm-up task) In the beginning of the experiment and at the very end, participants performed a simple melody (Appendix B.1) to the metronome click. To measure an effect of fatigue, a two-way ANOVA on

Fmean

with participants' professional

level (students vs. professionals) as between-subjects and pre-post condition as within-subjects factor was calculated. e ect of fatigue

The results revealed no signi cant

[F (1, 20) = 0.153, p = 0.7]

sional level of the players

and no signi cant e ect of profes-

[F (1, 20) = 0.443, p = 0.513],

nor any interaction .

The answers of the background questionnaire showed that the participants usually practise for a longer duration daily (M = 2.78 h, SD = 1.15) than the experiment lasted (approximately 1 h).

In uence of the sensor instrument to the measured nger forces In the questionnaire, participants evaluated the quality of the sensor equipped clarinet reed, the instrument, and also indicated whether they felt comfortable with playing this setup or not. The reed quality and the instrument quality had to be rated between +3 (very good) and 3 (very poor). Results showed, that the reed was evaluated as medium quality (M = 0.17, SD = 1.75) and the instrument as well (M = 0.74, SD = 1.1). We tested

Fmean

for the group

e ect of setup evaluation (players felt comfortable vs. players who reported issues with the setup) by separate between-subjects ANOVAs and found no signi cant e ect.

4.3

Discussion

This study investigated nger force pro les and temporal e ects of professional clarinetists and advanced clarinet students while performing on a sensor clarinet.

The measurement setup captured nger forces of index nger, middle

nger and ring nger of both hands, applied to the six main tone holes, as well as the reed oscillations during performance. The average nger forces participants used to close the tone holes during expressive performance was measured with

1 Data

Fmean = 1.21 N. Compared to nger

from two participants had to be removed, due to intermission of the sensor data.

Chapter 4. Finger forces in clarinet playing

forces reported for other instruments , this value is low. Even less force was used, when the participants had to perform a technical exercise task focussing on articulation techniques and timing (Fmean = 0.64 N). Even though, the average peak nger force captured during the entire study (Fmax = 12.95 N) is high in relation to the average nger force, it is still below the recommended nger force of 50 N recommended for proper holding a guitar string (Hori

et al., 2013). Hence, clarinet playing requires very low forces of index nger, middle nger and ring nger of both hands, compared to playing other musical instruments. Practising and performing a musical instrument over a long time is an unilateral stress for the body. Using too much tension during playing may result in overuse syndromes.

Force measurements have been shown to be a useful

tool to pinpoint reasons for overstressed body parts (Grosshauser and Troester, 2013). Although index nger, middle nger and ring nger have not primarily been reported as problematic areas of clarinetists (Diethelm, 2011), we hypothesised that using too much tension with these ngers may be an indicator for an overall uptightness during performance. Such an uptightness may lead to physical discomfort in relation to clarinet playing. However, the group of players who reported physical discomfort in relation to their clarinet playing, did not show signi cantly di erent nger force pro les compared to the group of players reporting no such problems. Consequently, measuring forces of these particular six ngers provides no reliable indicator to forecast overuse syndromes in clarinet playing. Additionally, when participants had to report how much nger force they think they use, they showed good results in their self-estimation. Measuring nger forces on the clarinet still remains a complicated procedure. Although the ring-shaped sensors were capable to measure nger forces applied to the six tone holes of the instrument, these sensors turned out to be fragile. Some sensors already broke during this study. Possible reasons might be a sliding from side keys to the tone holes and movements of the arms, as

1 Holding

down violin strings was measured with (Fmean >2.7 N, Kinoshita and Obata, 2009), guitar strings clamping force was (Fmean >30 N, Hori et al., 2013) and also the force required to play a grand piano in forte dynamics was measured with (Fmean >5 N, Parlitz et al., 1998).

Chapter 4. Finger forces in clarinet playing

reported in Wanderley et al. (2005). Such player actions cause a rotation of the ngers. The occurring rotational forces work against the construction principle of the ring shaped force sensors, which were designed to be pressed only from a perpendicular direction. In worst case, rotating the sensor disengaged the polymer ring from the glass balls (see Figure 4.1) or the sensing elements from the LTCC ring, which were glued with a conductive silver/epoxy adhesive (see Figure 5.20 in Weilguni, 2013). Once the glued electrical connection was broken, artefacts interfered the measurements. The results of the study showed that nger forces in clarinet playing are very low compared to nger forces on other musical instruments.

Further-

more, these forces seem not to be restricted to only vertical movements of the ngers. Attaching force sensors directly at the nger tips of the player, would additionally allow to measure nger forces of the same player performing on di erent types of clarinets, but also allows to use the setup to capture nger forces on other instruments, like saxophone or ute.

On wind instruments,

were key pads close the tone holes (e.g., on ute Fabre et al. 2012), it would be interesting to investigate in how far a leakage of the pads in uences the nger forces. For the technical exercise task of the study, we repeated a task from a previous saxophone study (Hofmann and Goebl, 2014). This allows to compare temporal e ects of clarinet performances with those on the saxophone. Participants performed a melody which was designed to be played by di erent e ectors ( ngers only, tongue only, tongue- nger actions), under three di erent tempi. Overall, clarinetists showed a higher timing precision in comparison to the saxophonists, especially with tongue-only actions. Nevertheless, similar trends were found in the results of both studies.

For the slow tempo con-

dition, combined tongue- nger actions achieved the highest timing precision. Although clarinetists' tongue-only actions showed the highest timing precision in all three tempo conditions, we again observed the e ect that for the fast tempo condition, combined tongue- nger actions were close to the CV of nger-only actions. This is yet another indicator, that ngers play a dominant role in the overall timing of woodwind performances. Taking the nger force measurements from the technical exercise into account, the study suggests that professional players are able to adjust their nger forces to very light ngering

Chapter 4. Finger forces in clarinet playing

technique (Fmean = 0.54 N) to achieve better timing.

Chapter 5

Application of performance measurements to physics based sound synthesis

In this chapter physical modelling sound synthesis is used to illustrate the di erence between tongue articulation and non-tongue articulation in singlereed woodwind performance. The parameter changes observed in the sensor reed measurements on saxophone and clarinet in Chapter 2.3 will be applied to a lumped mass-spring model (Chatziioannou and van Walstijn, 2012), in order to simulate the two di erent articulation techniques used by professional players.

5.1

Introduction

Physical modelling of acoustic instruments has various applications. On one hand, these models are used to explain the behaviour of real instruments (VĂ¤limĂ¤ki, 2004; Facchinetti et al., 2003; Dalmont et al., 2005), the interaction between player and instrument (Barthet et al., 2010; Van Walstijn and Avanzini, 2007; Almeida et al., 2013) or help to analyse the in uence of material properties to the radiated sound (Chatziioannou, 2010, see Chapter 6.6). On the other hand, physical modelling is a promising approach to synthesize sound for musical applications (Roads, 1996; Cook, 2002). Based on acoustical measurements on single-reed instruments (e.g., Boutil-

Chapter 5. Application of performance measurements to physics based sound

synthesis

lon and Gibiat, 1996; Nederveen, 1998; Dalmont et al., 2003; Gazengel et al., 2007), physical models for computer-based sound synthesis of steady-state tones have been developed (Chatziioannou, 2010; Karkar et al., 2012; Van Walstijn and Avanzini, 2007; Facchinetti et al., 2003).

Sounds, synthesized using a physical modelling technique can be shaped by changing the control parameters of the model. This allows to modulate the synthesized sound in a similar way as controlling a real instrument. Of particular interest is the control of the player over the instrument during expressive performance. The way players shape note onsets and note o sets during saxophone and clarinet performance is essential for a high level of expressiveness in their performance (see Chapter 2.1.1).

The embouchure of the player together with the reed-mouthpiece system can be described as a non-linear excitation mechanism that is coupled to a resonance tube (Chatziioannou and van Walstijn, 2012). During note onsets and note o sets in expressive woodwind performance, professional musicians use various articulation techniques to control the oscillations of the instrument. In e ect, transient phenomena arise at the beginning or at the end of a tone, or during note-to-note transitions. The underlying physics of these short transients is hard to measure under real playing conditions.

In the last years the interest in modelling transient behaviour for single-reed instruments grew (Sterling et al., 2009; Bergeot et al., 2012; Guillemain and Vergez, 2012). Two di erent assumptions of the tonguing related parameter changes were discussed earlier (Chapter 2.3, Hofmann et al. (2012a): Ducasse (2003) described the damping e ect of the tongue to the reed and its force to change the equilibrium position of the reed. In contrast to that, Sterling

et al. (2009) modelled the tongue as a gate to the mouthpiece, which prevents the air-stream to enter. The performance measurements presented in Chapter 2.3.3 and 3.2.2 showed that there is a clear distinction between the signals (reed bending, mouthpiece pressure, blowing pressure) captured for both articulation techniques. Following, the results of the performance observations will be used to derive control parameters for tongued and air-separated tone transitions, that can be applied to physical modelling sound synthesis.

Chapter 5. Application of performance measurements to physics based sound

synthesis

5.2

Physical modelling

In this experiment, the physical model of a clarinet, presented in Chatziioannou and van Walstijn (2012), will be adopted to simulate tongued tone transitions and air-separated tone transitions for the saxophone and the clarinet. This model consists of a lumped mass-spring oscillator, coupled to the impulse response of an appropriate resonator for saxophone (simple conical tube) or clarinet (straight cylindrical tube). The aim of this study is to model articulation by only modifying existing model parameters. No complexity will be added to the model.

In a rst step, new model parameters for steady-state

sounds were estimated through inverse modelling. Mouthpiece pressure measurements of a steady portion of the sound (see Chapter 2.3.3, Figure 2.12) were used for this process. Details about the procedure of estimating the model parameters for saxophone and clarinet can be found in the related publications (Chatziioannou and Hofmann, 2013, 2015)

5.2.1 Modelling articulation on single-reed woodwind instruments A rst propose on how to model tongue and non-tongue articulation on the saxophone was made in Hofmann et al. (2012a). Additional measurements on the clarinet (given in Chapter 2.3.3) and further simulations (Chatziioannou and Hofmann, 2013, 2015) were required to constitute these ndings.

The

following section summarizes the parameter settings that were extracted from these measurements and simulations and shows how they can be applied to physics based sound synthesis.

Tongue articulation To model a portato tongue stroke to the vibrating reed, a tongue-reed contact duration of 30 ms is used (see Chapter 3.2.2, Characteristics of articulation

techniques ). 1I

In reference to the observed e ects of the player's tongue to

am second author in both publications and contributed to the de nition of the articulation varied parameters and provided the processed data of the performance measurements. Vasileios Chatziianou programmed the physical model and developed the mathematical description, which is not part of this thesis.

Chapter 5. Application of performance measurements to physics based sound

1e+04 0.20 0.10 0.00 0.1 0.3

Reed displacement (mm)

0.5

Effective mass (kg)

1e+02

Damping (1/s)

1e+06

synthesis

0.00

0.05

0.10

0.15

Time (s)

Figure 5.1: Tongue-reed contact is modelled by increasing the damping (top), the e ective mass (middle), and the reed displacement (bottom) of the mass-spring oscillator for 30 ms. the vibrating sensor reed, described in Chapter 2.3.3 (Extracted parameters:

Tongued tone transition ), the following parameters of the physical model are adjusted:

â&#x20AC;˘

increase of the internal damping of the mass-spring oscillator

â&#x20AC;˘

increase of the e ective mass of the oscillator

â&#x20AC;˘

displacement of the equilibrium of the oscillator towards the mouthpiece

The blowing pressure remains approximately constant in this case, as veri ed by the measurements from Chapter 2.3.3 (Figure 2.13, left). The variation of the related model parameters is shown in Figure 5.1.

Air-separated tones The sensor reed measurements in Chapter 2.3.3 (Extracted parameters: Air-

separated tones ) showed that in the case of air-separated tone transitions, the reed vibrations were not directly manipulated by the player's tongue. To model air-separated tone transitions, only one parameter of the model is adjusted:

Chapter 5. Application of performance measurements to physics based sound

3.5 3.0 2.5 2.0

Blowing pressure (kPa)

4.0

synthesis

0.0

0.1

0.2

0.3

0.4

0.5

Time (s)

Figure 5.2: Air-separated tones tones are modelled by a modulation of the blowing pressure.

â&#x20AC;˘

reduction of blowing pressure

Figure 5.2 shows the variation of the blowing pressure parameter of the model.

5.2.2 Results1 The simulated pressure and reed displacement signals are plotted in Figure 5.3 and Figure 5.4, showing both modelled articulation techniques. The presented model parameters make a clear distinction between the two articulations. The synthesized signals show similarities with the measurements presented earlier in Chapter 2.3. In particular, in both Figures concerning the clarinet, Figure 2.13 (measurement, page 27) and Figure 5.3 (model) (Figure 2.14 and 5.4 for the saxophone), the modulation of the reed oscillations were softer when the tongue was used, because of the direct but short interaction with the reed.

In the case of the model, the reed displacement was

calculated (Figure 5.3, top), whereas during the experiments the bending of the reed was measured (Figure 2.13, top). These two signals are not directly comparable. As a consequence, the measurements can only be used in a qualitative prediction of the motion of the reed. Also for the mouthpiece pressure, a characteristic pressure envelope for each articulation technique was visible in the measured signals. This `articulation

1 as

published in Hofmann et

al.

(2013a); Chatziioannou and Hofmann (2013, 2015)

Chapter 5. Application of performance measurements to physics based sound

synthesis

signature' is also present in the simulated pressure signals of Figure 5.3 (bottom). Finally, the increase in the reed damping caused by the contact with the tongue results in a smoother synthesised pressure signal during the transient, in comparison to the case of air-separated tones. Such an e ect has also been observed in the experimental measurements (see Figure 5.1). ples generated by the model are online available at:

Sound exam-

http://iwk.mdw.ac.at/

?page_id=148&sprache=2.

5.3

Discussion

Sensor-based measurements during professional woodwind performance are a useful tool to collect parameter variations to control physical models. Understanding the player-instrument interactions helps to synthesize more realistic sounds. Capturing player-instrument interactions during performance always involves limitations in the amount of parameters to track.

It is a matter of

nding the right balance between adding sensor technology to the acoustic instrument and not distracting the player. Therefore observations are limited to a speci c parameter set. From the obtained blowing pressure, reed bending and mouthpiece pressure measurements (Chapter 2.3), useful variations of articulatory controlled parameters for the model were derived.

An application of these parameter

changes to the model allowed to reproduce the measured pressure signals, at least qualitatively, for steady-state and during transient oscillations (Hofmann

et al., 2012a; Chatziioannou and Hofmann, 2013). In future work this approach will be enhanced towards the simulation of numerous articulation techniques. To develop a more robust re-synthesis routine, the reed bending signal needs to be directly correlated with reed displacement. Furthermore, to ensure the quality of the re-synthesized sounds, listening tests may be used as a measurement tool. These tests can range from sound comparisons between real recordings and synthetic sounds, to discrimination tasks within the pool of synthesized sounds. If participants are able to discriminate synthesized articulation techniques with the same precision as with recorded sound samples, this would verify the variation of the model parameters.

Chapter 5. Application of performance measurements to physics based sound

0.05

0.10

4e−04 2e−04 0e+00 4 2 0 −2

Mouthpiece pressure (kPa) 0.00

−4

Reed displacement (m)

4e−04 2e−04 0e+00 4 2 0 −2 −4

Mouthpiece pressure (kPa)

Reed displacement (m)

synthesis

0.15

0.00

0.05

Time (s)

0.10

0.15

Time (s)

Figure 5.3: Simulated reed displacement (top) and resulting mouthpiece pressure (bottom) for a single note transition on the clarinet. Modelling results of tongue separated

0.15

2e−04 0e+00 1.5 0.5 −0.5

Mouthpiece pressure (kPa) 0.10

−1.5

Reed displacement (m)

2e−04 0e+00 1.5 0.5 −0.5 −1.5

Mouthpiece pressure (kPa)

Reed displacement (m)

tones (left) and air-separated tones (right).

0.20

Time (s)

0.10

0.15

0.20

0.25

Time (s)

Figure 5.4: Simulated reed displacement (top) and resulting mouthpiece pressure (bottom) for a single note transition on the saxophone. Modelling results of tongue separated tones (left) and air-separated tones (right).

Chapter 5. Application of performance measurements to physics based sound

100

synthesis

Chapter 6

Conclusion and future work

The research presented in this thesis investigated the interaction between musician and acoustic instrument in saxophone and clarinet performance.

The

speci c focus was tongue articulation measurements and nger force measurements on the saxophone and the clarinet. Within this dissertation, a unique method for capturing reed oscillations was developed.

The main objective was to monitor tongue-reed interaction

during performance. Reed modi cations, sensor calibration, and preliminary tests are presented in Chapter 2. Although the sensor reeds were primarily designed to track tongue articulation during empirical performance studies with saxophone and clarinet, they also created precise measurements of articulation parameters, showing the e ects of the tongue to the reed. Three empirical performance studies (two production experiments and one perception experiment) make up the majority of this thesis (Chapter 3 and 4). The corpus of data within the production experiments contains more than 40 hours of saxophone and clarinet performances.

Sensor reeds were used

to monitor tongue articulation. To measure nger forces in clarinet playing, Weilguni (2013) developed special ring-shaped force sensors for a Viennese Clarinet at the Vienna Technical University. As a nal step, the measured physical parameters of professional players' articulation techniques were applied to a physical model in an attempt to simulate transient behaviour for tone transitions (Chapter 5).

101

102

6.1

Chapter 6. Conclusion and future work

Summary of contributions

The main contributions of this thesis are:

â&#x20AC;˘

A measurement setup to capture reed oscillations during performance was developed. After calibrating the reed sensor, it was possible to extract tonguing parameters from professional players. The observations support a tonguing model proposed by Ducasse (2003), in which a tongue stroke shows a damping e ect to the vibrating reed, and its force changes the equilibrium position of the reed during a tongued tone transition.

â&#x20AC;˘

Through simultaneous measurements of blowing pressure, mouthpiece pressure, and reed bending, physical control parameters for tongued and non-tongued articulation were gained.

These physical control parame-

ters were applied to a physical model (Chatziioannou, 2010) to simulate transient behaviour in tone transitions. The reproduced pressure signals of the model showed a clear distinction between the two articulation techniques.

â&#x20AC;˘

Considering that saxophonists have to coordinate tongue and nger actions during expressive performance, an empirical study (N = 19) to examine temporal e ects of di erent e ector combinations was carried out. Through the sensor reed measurements, it was possible to distinguish between tongued and non-tongued tones in the recorded sequences. For slow tempi, the performance timing bene ted from combined tongue nger actions. This supports the multiple-timer model from Ivry et al. (2002), which predicts stabilized timing for combined e ectors. However, in the fast tempo condition, nger actions showed a dominant e ect on the overall timing (Hofmann and Goebl, 2014).

â&#x20AC;˘

A reproduction of the saxophone experiment with clarinetists (N = 23) con rmed this observation: In the fast tempo condition, nger actions play a dominant role on the timing. Even tough in the case of the clarinetists, tongue-only timing was superior than that of the saxophonists, combined tongue- nger actions followed the timing of the ngers.

â&#x20AC;˘

In a listening study (N = 31), it was tested if expertise in saxophone

103

Chapter 6. Conclusion and future work

playing helps to facilitate the discrimination of articulation techniques. Participants with a di erent background in music making attended the experiment (saxophonists, musicians not playing a wind instrument, nonmusicians).

Independent of the listener's group, errors occurred when

legato articulation and portato articulation had to be discriminated, nevertheless saxophonists showed the best results for this task. In reference to the motor theory of speech perception (Galantucci et al., 2006), the results of the experiment indicate that the link between production and perception of speech may also apply to the production and perception of articulation in saxophone music (Hofmann and Goebl, 2014).

â&#x20AC;˘

To examine nger forces with clarinet students (N = 17) and professional clarinetists (N = 6), an empirical study comprised of a variety of performance conditions was designed and conducted. In comparison to nger forces reported for other instruments (Kinoshita and Obata, 2009; Hori et al., 2013; Parlitz et al., 1998), the measured average nger forces for expressive playing was low (Fmean = 1.21 N). Finger forces measured during a technical exercise task was even lower (Fmean = 0.64 N). The group of professional clarinetists showed very light nger forces to achieve precise timing in the technical exercise (Fmean = 0.54 N). These observations indicate that professional players use very light ngering technique if possible.

6.2

Future work

The sensor instruments developed and used in this research enabled an investigation of ne motor control of clarinet and saxophone players. There are some technical aspects with the sensor technology that can be optimised in future research. During the nger force measurements on saxophone and clarinet, it turned out that both setups showed weaknesses.

On the saxophone, the standard

industry foil sensors attached to the pearl buttons were not touched by all players.

Here, it might be a possibility to place the sensor inside the key

between the pad and the cup. The ring shaped force sensors on the clarinet

104

Chapter 6. Conclusion and future work

showed two weaknesses. First, the polymer rings glued on three glass balls fell o . Second, the electric conductive silver/epoxy adhesive connection between the sensing elements and the Low Temperature Co- red Ceramic (LTCC) ring broke in some cases. The developed sensor reed has only one strain gauge sensor attached. This limits the observations to only one speci c area of the reed. For example, this does not allow for the observation of torsional modes that may occur in the vibration patterns of clarinet reeds, as reported in Pinard et al. (2003). Using a high-speed camera and an arti cial blowing machine for further calibration would help to monitor the tip of the sensor reed. The position and the amount of strain gauge sensors on the reed could then be optimized. Furthermore, it provides details about the link between reed bending measurements and the tip opening. Compared to more traditional synthesis methods that attempt to replicate sound spectra, physical modelling aims to simulate the sound generation mechanism of musical instruments. This o ers a more realistic reproduction of waveforms and also allows to include player actions that are controlling the involved physical parameters.

With the link between reed bending and tip

opening, it would be possible to directly compare the behaviour of a physical model with the measured sensor reed signal, captured in a human performance. This could be a way to further optimize the physical model towards realistic tone transitions, which are capable of simulating a variety of articulation techniques, similar to the repertoire of professional players. Another way to verify sounds of a model, apart from comparing physical parameters, could be listening tests. Similar, to the AB-X test presented in Chapter 3.3, the modelled sounds could be in a pool of stimuli. If participants can discriminate modelled articulation techniques with the same precision as recorded articulation techniques, this would additionally verify the model parameters applied for the tone transitions. The combination of measurement techniques from music acoustics and the empirical approach from performance science research would allow a comparison of style between jazz and classical soloists. Applying stylistically grouped control parameters to a physical model, may be the key to re-synthesize more expressive tone transitions in comparison to traditional sample playback.

Chapter 6. Conclusion and future work

105

Apart from looking only at physical parameter changes in solo performances, such sensor instruments can be used to investigate a large variety of di erent performance situations.

A future application may be to investigate the role

of each musician in a music ensemble. Especially for the saxophone, which is primarily played in jazz music, this might be an approach to better understand groove in jazz ensembles (Prรถgler, 1995). A combination of sensor instruments, video technology, and acceleration sensors, also on the musician's body, gives the opportunity to study ensemble synchronisation at various levels, ranging from large pick-up gestures to the ne motor control of each individual player.

106

Chapter 6. Conclusion and future work

Bibliography

2009).

Almeida, A., Chow, R., Smith, J., and Wolfe, J. (

The kinetics and

acoustics of ngering and note transitions on the ute , The Journal of the Acoustical Society of America

126, 1521 1529. 2013). The clarinet:

Almeida, A., George, D., Smith, J., and Wolfe, J. (

How

blowing pressure, lip force, lip position and reed hardness a ect pitch, sound level, and spectrum , The Journal of the Acoustical Society of America

134, 2247 2255. Almeida, A., Lemare, J., Sheahan, M., Judge, J., Auvray, R., Dang, K., John,

2010). Clarinet parame-

S., Geo roy, J., Katupitiya, J., Santus, P., et al. (

ter cartography: automatic mapping of the sound produced as a function of blowing pressure and reed force , in Proceedings of the International Sym-

posium on Music Acoustics (Sydney, Australia).

2004). Chronische Schmerzen beim Mu-

AltenmĂźller, E. and Jabusch, H.-C. (

sizieren , Das Orchester 7/8 17 22.

Baader, A., Kazennikov, O., and Wiesendanger, M. (

2005).

Coordination

of bowing and ngering in violin playing , Cognitive Brain Research

23,

436 443.

2010).

Barthet, M., Guillemain, P., Kronland-Martinet, R., and Ystad, S. (

From clarinet control to timbre perception , Acta Acustica united with Acustica

Bates, P. (

96, 678 689.

1984). The New Grove Dictionary of Musical Instruments, chapter

Tonguing, edited by Sadie, S. (London: Macmillan Press Ltd.).

107

108

BIBLIOGRAPHY

1983). Analysis and synthesis of musical rhythm , Studies of music performance 39, 27 60.

Bengtsson, I. and Gabrielsson, A. (

Bergeot, B., Almeida, A., Vergez, C., and Gazengel, B. (

2012). Attack tran-

sients in a clarinet model with time-varying blowing pressure , in Proceedings

of the Acoustics 2012 Nantes Conference, 1981 1986 (Nantes, France). Bertsch, M. and Mayer, A. (

2005). A new 3d transducer for measuring the

trumpet mouthpiece force , in Proceedings of the Forum Acusticum, 23 30 (Budapest, Hungary).

2009). ...in Liebe zer ossenes GefĂźhl Die Klarinette, Symposium

Birsak, K. (

im Rahmen der 30. Tage Alter Musik in Herne 2005, chapter 'Zur Geschichte des vollen Klarinettentones', 42 63 (MĂźnchen: Musikverlag Katzbichler).

2014). Context-speci c e ects of musical expertise integration , Frontiers in Psychology: Cognitive Science 5,

Bishop, L. and Goebl, W. ( on audiovisual 1 14.

1996).

Boutillon, X. and Gibiat, V. (

Evaluation of the acoustical sti ness

of saxophone reeds under playing conditions by using the reactive power approach , The Journal of the Acoustical Society of America

100,

1178

1189.

2006). The aubio library at mirex 2006 , Music Information

Brossier, P. M. (

Retrieval Evaluation eXchange 2006 1 3.

2012). Segmenting and parsing instrumentalists' gestures , Journal of New Music Research 41, 13

Caramiaux, B., Wanderley, M., and Bevilacqua, F. (

29.

Chatziioannou, V. (

2010).

Forward and inverse modelling of single-reed

woodwind instruments with application to digital sound synthesis , Ph.D. thesis, Queen's University Belfast.

Chatziioannou, V. and Hofmann, A. (

2013). Modeling articulation techniques

in single-reed woodwind instruments , in Proceedings of Meetings on Acous-

tics, volume 19, 035059 (ASA, Montreal, Canada).

109

BIBLIOGRAPHY

Chatziioannou, V. and Hofmann, A. (

2015). Physics-based analysis of artic-

ulatory player actions in single-reed woodwind instruments , Acta Acustica united with Acustica

101, 292 299. 2012). Estimation of clarinet reed Acta Acustica united with Acustica 98,

Chatziioannou, V. and van Walstijn, M. ( parameters by inverse modelling , 629 639.

Chen, J., Smith, J., and Wolfe, J. (

2011). Saxophonists tune vocal tract reso-

nances in advanced performance techniques , The Journal of the Acoustical Society of America

129, 415 426.

2002). Real Sound Synthesis for Interactive Applications (Welles-

Cook, P. R. (

ley: AK Peters, Ltd.).

2004). Playing the accent - comparing striking velocity and timing

Dahl, S. (

in an ostinato rhythm performed by four drummers , Acta Acustica united with Acustica

90, 762 776.

Dahl, S., Bevilacqua, F., Bresin, R., Clayton, M., Leante, L., Poggi, I.,

2009). Musical Gestures:

and Rasamimanana, N. (

Sound, Movement, and

Meaning, chapter Gestures in performance, 36 68 (New York: Routledge).

2003). Nonlinear characteristics of

Dalmont, J., Gilbert, J., and Ollivier, S. (

single-reed instruments: Quasistatic volume ow and reed opening measurements , The Journal of the Acoustical Society of America

114, 2253 2262. 2005). An ana-

Dalmont, J.-P., Gilbert, J., Kergomard, J., and Ollivier, S. (

lytical prediction of the oscillation and extinction thresholds of a clarinet , The Journal of the Acoustical Society of America

118, 3294 3305.

D'Ausilio, A., PulvermĂźller, F., Salmas, P., Bufalari, I., Begliomini, C., and Fadiga, L. ( Biology

2009). The motor somatotopy of speech perception , Current

19, 381 385.

Desmet, F., Nijs, L., Demey, M., Lesa re, M., Martens, J.-P., and Leman,

2012). Assessing a clarinet player's performer gestures in relation to locally intended musical targets , Journal of New Music Research 41, 31 48. M. (

110

BIBLIOGRAPHY

Diethelm,

2011).

(

Schmerzen

Gesellschaft für Musik-Medizin

beim

Musizieren ,

Schweizerische

5, 60.

2003). A physical model of a single-reed wind instrument, including actions of the player , Computer Music Journal 27, 59 70.

Ducasse, E. (

Fabre, B., Auvray, R., Curtit, M., and Roosen, J.-Y. (

2012). Measuring leak-

ages in ute pads , in Proceedings of the Acoustics 2012 Nantes Conference (Nantes, France). Facchinetti, M., Boutillon, X., and Constantinescu, A. (

2003). Numerical and

experimental modal analysis of the reed and pipe of a clarinet , The Journal of the Acoustical Society of America

113, 2874 2883.

Fadiga, L., Craighero, L., Buccino, G., and Rizzolatti, G. (

2002).

Speech

listening speci cally modulates the excitability of tongue muscles: study , European Journal of Neuroscience

a tms

15, 399 402.

1997). Neurobiology of speech perception , Annual Review of Neuroscience 20, 331 353.

Fitch, R. H., Miller, S., and Tallal, P. (

1979).

Fletcher, N. H. (

Air ow and sound generation in musical wind in-

struments , Annual Review of Fluid Mechanics

11, 123 146.

Forssberg, H., Eliasson, A., Kinoshita, H., Johansson, R., and Westling, G.

1991).

(

Development of human precision grip I: Basic coordination of

force , Experimental Brain Research

85, 451 457.

2003). Music performance research at the millennium , Psychology of music 31, 221 272.

Gabrielsson, A. (

2006). The motor theory of speech perception reviewed , Psychonomic Bulletin & Review 13, 361 377.

Galantucci, B., Fowler, C. A., and Turvey, M. T. (

2010). Ultrasonic investigation of clarinet multiple articulation ,

Gardner, J. (

Ph.D. thesis, Arizona State University. Gazengel, B., Guimezanes, T., Dalmont, J.-P., Doc, J., Fagart, S., and Léveillé,

2007).

Y. (

Experimental investigation of the in uence of the mechanical

111

BIBLIOGRAPHY

characteristics of the lip on the vibrations of the single reed , in Proceedings

of the International Symposium on Musical Acoustics (Barcelona, Spain). Gingras, B. and McAdams, S. (

2011).

Improved score-performance match-

ing using both structural and temporal information from midi recordings , Journal of New Music Research

40, 43 57.

2008). Tactile feedback and timing accuracy in piano performance , Experimental Brain Research 186, 471 479.

Goebl, W. and Palmer, C. (

2009).

Finger motion in piano performance:

Touch and tempo , in International Symposium on Performance Science

(ISPS), 65 70 (Auckland, New Zealand).

2013).

Goebl, W. and Palmer, C. (

Temporal control and hand movement

e ciency in skilled music performance , PLoS ONE

1997).

Goolsby, T. (

8, e50901.

Verbal instruction in instrumental rehearsals: A com-

parison of three career levels and preservice teachers , Journal of Research in Music Education

45, 21 40. 2013). Automatic align-

Grachten, M., Gasser, M., Arzt, A., and Widmer, G. (

ment of music performances with structural di erences. , in Proceedings of

the 14th International Society for Music Information Retrieval Conference, 607 612 (Curitiba, Brazil). Grosshauser, T. and Troester, G. (

2013).

Sensor based hand weight and

pressure measurements in trombone playing , in Proceedings of the Stock-

holm Music Acoustics Conference 2013, SMAC 13, edited by R. Bresin and A. Askenfeldt, 398 402 (KTH Royal Institute of Technology, Stockholm, Sweden).

2012). Decay transients in single reed wind

Guillemain, P. and Vergez, C. (

instruments , in Proceedings of the Acoustics 2012 Nantes Conference, 1975 1979 (Nantes, France).

2010).

Guillemain, P., Vergez, C., Ferrand, D., and Farcy, A. (

An instru-

mented saxophone mouthpiece and its use to understand how an experienced musician plays , Acta Acustica united with Acustica

96, 622 634.

112

Hall, D. (

BIBLIOGRAPHY

1991).

Musical acoustics, Physics Series (California: Brooks/Cole

Publishing).

2001).

Haueisen, J. and Knรถsche, T. R. (

Involuntary motor activity in pi-

anists evoked by music perception , Journal of Cognitive Neuroscience

13,

786 792.

Hofmann, A., Chatziioannou, V., Kausel, W., Goebl, W., Weilguni, M., and

2012a). The in uence of tonguing on tone production with

Smetana, W. (

single-reed instruments , in Proceedings of the 5th Congress of the Alps Adria

Acoustics Association (Zadar, Croatia). Hofmann, A., Chatziioannou, V., Weilguni, M., Goebl, W., and Kausel, W.

2013a). Measurement setup for articulatory transient di erences in wood-

(

wind performance , in Proceedings of Meetings on Acoustics, volume 19, 035060 (Montreal, Canada).

2014).

Hofmann, A. and Goebl, W. (

Production and perception of legato,

portato and staccato articulation in saxophone playing , Frontiers in Psychology: Cognitive Science

5, 1 10. 2013b). Evaluating a wavelet-

Hofmann, A., Goebl, W., and Weilguni, M. (

based analysis of sensor reed signals for performance research , in Proceedings

of the Stockholm Music Acoustics Conference 2013, SMAC 13, edited by R. Bresin and A. Askenfeldt, 398 402 (KTH Royal Institute of Technology, Stockholm, Sweden).

Hofmann, A., Goebl, W., Weilguni, M., Mayer, A., and Smetana, W. (

2012b).

Measuring tongue and nger coordination in saxophone performance , in

12th International Conference on Music Perception and Cognition (ICMPC) and 8th Triennial Conference of the European Society for the Cognitive Sciences of Music (ESCOM), 442 445 (Thessaloniki, Greece). Hofmann, A., Goebl, W., Weilguni, M., and Smetana, W. (

2013c). Zooming

into saxophone performance: Tongue and nger coordination , in Proceed-

ings of the International Symposium on Performance Science, edited by A. Williamon and W. Goebl, 289 294 (Brussels, Belgium).

113

BIBLIOGRAPHY

Hori, M., Hosono, M., Takahashi, H., Matsumoto, K., and Shimoyama, I.

2013).

(

3-axis ngertip force during playing the string instrument , in

Proceedings of 2013 Transducers and Eurosensors XXVII: The 17th International Conference on Solid-State Sensors, Actuators, and Microsystems, 2745 2748 (IEEE, Barcelona, Spain).

Ivry, R., Richardson, T., and Helmuth, L. (

2002).

Improved temporal sta-

bility in multie ector movements. , Journal of Experimental Psychology: Human Perception and Performance

28, 72 92.

2013). Artikulation am Saxophon , (unpublished diploma the-

JaurovĂĄ, D. (

sis), UniversitĂ¤t fĂźr Musik und darstellende Kunst, Wien.

2010). Musical Ges-

Jensenius, A., Wanderley, M., GodĂ¸y, R., and Leman, M. (

tures: Sound, Movement, and Meaning, chapter Musical gestures: concepts and methods in research, 12 35 (New York: Routledge).

2012).

Karkar, S., Vergez, C., and Cochelin, B. (

Oscillation threshold of

a clarinet model: a numerical continuation approach , The Journal of the Acoustical Society of America

Kergomard, J. (

1995).

131, 698 707.

Mechanics of Musical Instruments, chapter Elemen-

tary considerations on reed-instrument oscillations, 229 290 (New York: Springer).

Kinoshita, H. and Obata, S. (

2009).

Left hand nger force in violin play-

ing: Tempo, loudness, and nger di erences , The Journal of the Acoustical Society of America

126, 388 395.

Kinoshita, H.and Obata, S., Hirano, T., Ohsawa, C., and Ito, T. (

2012).

Expert-novice di erence in string clamping force in violin playing , in 12th

International Conference on Music Perception and Cognition (ICMPC) and 8th Triennial Conference of the European Society for the Cognitive Sciences of Music (ESCOM), 532 533 (Thessaloniki, Greece).

1989). Schule fĂźr Klarinette in B: Deutsches und franzĂśsisches Sys-

Koch, E. (

tem (Leipzig: Deutscher Verlag fĂźr Musik).

114

BIBLIOGRAPHY

1982). Untersuchungen zur Artikulation bei Klarinettenin-

Krautgartner, K. (

strumenten im Jazz , Ph.D. thesis, UniversitĂ¤t zu KĂśln.

1987). Analysis of

Kronland-Martinet, R., Morlet, J., and Grossmann, A. (

sound patterns through wavelet transforms , International Journal of Pattern Recognition and Arti cial Intelligence

Legere, G. (

2000).

1, 273 302.

Oriented polymer reeds for musical instruments , US

Patent 6,087,571.

Legere, G., Kortschot, M., et al. (

2011). Oriented polymer reeds for woodwind

instruments , US Patent 7,902,443.

2010). Musical gestures:

Leman, M. and GodĂ¸y, R. (

Sound, movement, and

meaning, chapter 'Why Study Musical Gestures?', 3 11 (New York: Routledge).

1985). The motor theory of speech

Liberman, A. M. and Mattingly, I. G. ( perception revised , Cognition

21, 1 36.

1989). Developing a personal saxophone sound (Med eld:

Liebman, D. (

Dorn

Publications).

Manto, M., Bower, J. M., Conforto, A. B., Delgado-GarcĂa, J. M., da Guarda, S. N. F., Gerwig, M., Habas, C., Hagura, N., Ivry, R. B., MariĂŤn, P., et al.

2012). Consensus paper:

(

Roles of the cerebellum in motor control the

diversity of ideas on cerebellar involvement in movement , The Cerebellum

11, 457 487. 2011). School of Velocity (Schule der GelĂ¤u gkeit) (Mainz:

Mauz, R. (

Schott

Music).

Michaels, J. (

1999).

Methodische Schule der klarinettistischen Gri technik

(Frankfurt: Zimmermann Verlag).

Nederveen, C. (

1998). Acoustical Aspects of Woodwind Instruments

Northern Illinois University Press).

(DeKalb:

115

BIBLIOGRAPHY

Nemoto, K. and Arino, H. (

2007).

Hand and upper extremity problems in

wind instrument players in military bands , Medical Problems of Performing Artists

22, 67 69. 2007).

Palmer, C., Carter, C., Koopmans, E., and Loehr, J. (

Movement,

planning, and music: Motion coordinates of skilled performance , in Pro-

ceedings of the International Conference on Music Communication Science, 119 122 (Sydney, Australia).

2009a).

Palmer, C., Koopmans, E., Carter, C., Loehr, J., and Wanderley, M. (

Synchronization of motion and timing in clarinet performance , in Interna-

tional Symposium on Performance Science (Auckland, New Zealand).

2009b).

Palmer, C., Koopmans, E., Loehr, J., and Carter, C. (

Movement-

related feedback and temporal accuracy in clarinet performance , Music Perception

26, 439 449. 2006). Use of continuous wavelet-

Paradzinets, A., Harb, H., and Chen, L. (

like transform in automated music transcription , in Proceedings of the Eu-

ropean Signal Processing Conference (Florence, Italy). Parlitz, D., Peschel, T., and AltenmĂźller, E. (

1998). Assessment of dynamic

nger forces in pianists: e ects of training and expertise , Journal of Biomechanics Pay, A. (

31, 1063 1067.

1995). The Cambridge Companion to the Clarinet, chapter The me-

chanics of playing the clarinet, 107 122 (Cambridge: Cambridge University Press).

2006). Wavelet Methods for Time Series Anal-

Percival, D. and Walden, A. (

ysis, Cambridge Series In Statistical And Probabilistic Mathematics (Cambridge: Cambridge University Press).

2003).

Pinard, F., Laine, B., and Vach, H. (

Musical quality assessment of

clarinet reeds using optical holography , The Journal of the Acoustical Society of America PrĂśgler, J. A. (

113, 1736 1742.

1995). Searching for swing:

Participatory discrepancies in the

jazz rhythm section , Ethnomusicology 21 54.

116

BIBLIOGRAPHY

2012). R:

R Core Team (

A Language and Environment for Statistical Com-

puting, R Foundation for Statistical Computing, Vienna, Austria.

1993).

Rabiner, L. and Juang, B. (

Fundamentals of Speech Recognition,

Prentice-Hall Signal Processing Series: Advanced monographs (Englewood Cli s: Prentice Hall).

2012).

Rasamimanana, N. (

Towards a conceptual framework for exploring

and modelling expressive musical gestures , Journal of New Music Research

41, 3 12. 2005). Sensorimotor synchronization: a review of the tapping literature , Psychonomic Bulletin & Review 12, 969 992.

Repp, B. H. (

1996). The computer music tutorial

Roads, C. (

(Massachusetts: MIT Press).

2008). Measurement of vocal-

Scavone, G., Lefebvre, A., and da Silva, A. (

tract in uence during saxophone performance , The Journal of the Acoustical Society of America

123, 2391 2400. 2009).

Schoonderwaldt, E. and Demoucron, M. (

Extraction of bowing pa-

rameters from violin performance combining motion capture and sensors , The Journal of the Acoustical Society of America

1995).

Shackleton, N. (

126, 2695 2708.

The Cambridge Companion to the Clarinet, chapter

The Development of the Clarinet, 16 32 (Cambridge: Cambridge University Press).

2011).

Spahn, C., Richter, B., and AltenmĂźller, E. (

MusikerMedizin:

Di-

agnostik, Therapie und PrĂ¤vention von musikerspezi schen Erkrankungen (Stuttgart: Schattauer). Sterling, M., Dong, X., and Bocko, M. (

2009). Pitch bends and tonguing ar-

ticulation in clarinet physical modeling synthesis , in Proceedings of the 2009

IEEE International Conference on Acoustics, Speech and Signal Processing, 289 292 (IEEE Computer Society, Taipei, Taiwan). Tubiana, R. et al. (

2005). Functional anatomy of the hand / Movements of

the ngers / Fundamental positions for instrumental musicians (anniversary article) , Medical Problems of Performing Artists

20, 183.

117

BIBLIOGRAPHY

2001).

Tzanetakis, G., Essl, G., and Cook, P. (

Audio analysis using the

discrete wavelet transform , in Proceedings of the Conference in Acoustics

and Music Theory Applications, 318 323 (Skiathos, Greece).

2004). Physics-based modeling of musical instruments , Acta Acustica united with Acustica 90, 611 617.

VĂ¤limĂ¤ki, V. (

2007). Modelling the mechanical response

Van Walstijn, M. and Avanzini, F. (

of the reed-mouthpiece-lip system of a clarinet. Part II: A lumped model approximation , Acta Acustica united with Acustica

93, 435 446.

2014). Adaptive calibration of a three-

van Walstijn, M. and de Sanctis, G. (

microphone system for acoustic waveguide characterization under timevarying conditions , The Journal of the Acoustical Society of America

135,

917 927. Wanderley, M., Vines, B. W., Middleton, N., Mckay, C., and Hatch, W.

2005). The musical signi cance of clarinetists' ancillary gestures: ploration of the eld , Journal of New Music Research 34, 97 113. (

Wehle, R. (

2007).

An ex-

Clarinet Fundamentals Volume 2: Systematic Fingering

Course, Clarinet Fundamentals (Mainz: Schott Music). Weilguni, M. (

2013).

Force sensors for the measurement of nger forces in

clarinet playing , Ph.D. thesis, Vienna University of Technology.

2013).

Weilguni, M., Hofmann, A., Smetana, W., and Goebl, W. (

Fabri-

cation and mechanical characterization of ring-shaped LTCC/HTCC nger force sensors used for performance studies in clarinet playing , in The 17th

International Conference on Solid-State Sensors, Actuators and Microsystems (Barcelona, Spain). Weilguni, M., Smetana, W., Hofmann, A., Radosavljevic, G., Goebl, W., and Nicolics, J. (

2012).

A ring-shaped LTCC/HTCC sensor for detection of

nger forces in clarinet playing , in IEEE Sensors 2012 Conference (Taipei, Taiwan).

1992). Strain Gauge Technology, An Objective publication (New

Window, A. (

York: Springer).

118

BIBLIOGRAPHY

2007). When the brain plays

Zatorre, R. J., Chen, J. L., and Penhune, V. B. ( music:

auditory motor interactions in music perception and production ,

Nature Reviews Neuroscience

8, 547 558.

Appendix A

Additional material saxophone experiments

119

120

Chapter A. Additional material saxophone experiments

Note-onsets (TRR)

Note-o sets (TRC)

Prec.

Rec.

F-Meas.

Prec.

Rec.

F-Meas.

Portato Slow

0.9793

0.9861

0.9827

0.9862

0.9931

0.9896

Portato Med.

0.8767

0.8828

0.8797

0.9247

0.9310

0.9278

Portato Fast

0.9861

0.9726

0.9793

0.9653

0.9521

0.9586

Staccato Slow

0.9388

0.9517

0.9452

0.9456

0.9586

0.9521

Staccato Med.

0.9728

0.9795

0.9761

0.9864

0.9932

0.9898

Staccato Fast

1.0000

0.9384

0.9682

0.9854

0.9310

0.9574

Legato Slow

0.9390

0.9625

0.9506

Legato Med.

0.9792

0.9691

0.9741

Legato Fast

0.9452

0.6900

0.7977

0.958

0.931

0.944

0.965

0.960

0.963

Sum

Table A.1:

F-measure, precision

and

recall

for wavelet based onset and o set de-

tection in the sensor reed signal. Results for a Âą25 ms evaluation window are given and further discussed in Section 3.2.1.

121

Chapter A. Additional material saxophone experiments

IOI − Similarity

Pitch − Similarity

10 5

0 0 0 0 0 0 0 0 0 0 0.020.030.050.110.29 1 0 0 0 0 0 0 0 0 0 0.020.030.050.110.29 1 0.29 0 0 0 0 0 0 0 0 0.020.030.050.110.29 1 0.290.11 0 0 0 0 0 0 0 0.020.030.050.110.29 1 0.290.110.05 0 0 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.03 0 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0.020.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0.030.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0.050.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0 0.110.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0 0 0.29 1 0.290.110.050.030.02 0 0 0 0 0 0 0 0 0 1 0.290.110.050.030.02 0 0 0 0 0 0 0 0 0 0

Performance Events

10 5

Performance Events

1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1 0.050.050.050.050.110.29 1 0.050.110.29 1 0.050.110.29 1 0.05 0.110.110.110.110.29 1 0.290.110.29 1 0.290.110.29 1 0.290.11 0.290.290.290.29 1 0.290.110.29 1 0.290.110.29 1 0.290.110.29 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1 0.050.050.050.050.110.29 1 0.050.110.29 1 0.050.110.29 1 0.05 0.110.110.110.110.29 1 0.290.110.29 1 0.290.110.29 1 0.290.11 0.290.290.290.29 1 0.290.110.29 1 0.290.110.29 1 0.290.110.29 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1 0.050.050.050.050.110.29 1 0.050.110.29 1 0.050.110.29 1 0.05 0.110.110.110.110.29 1 0.290.110.29 1 0.290.110.29 1 0.290.11 0.290.290.290.29 1 0.290.110.29 1 0.290.110.29 1 0.290.110.29 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1 1 1 1 1 0.290.110.05 1 0.290.110.05 1 0.290.110.05 1

Score Events

Final Cost Matrix d)

10 Score Events

●

−1

●

−2

●

1 Reference index

1 1 1 1 1 1 1 1 1 1 1 0.970.980.990.98 0 1 1 1 1 1 1 1 1 1 0.990.97 1 0.990.91 0 0.98 1 1 1 1 1 1 1 1 0.990.970.980.990.91 0 0.910.99 1 1 1 1 1 1 1 0.990.970.980.990.91 0 0.910.990.98 1 1 1 1 1 1 1 0.970.980.990.98 0 0.910.99 1 0.97 1 1 1 1 1 0.990.97 1 0.990.91 0 0.980.990.980.97 1 1 1 1 1 0.990.970.980.990.91 0 0.910.990.980.970.99 1 1 1 1 0.990.970.980.990.91 0 0.910.990.980.970.99 1 1 1 1 0.980.970.980.990.98 0 0.910.99 1 0.970.99 1 1 1 1 1 1 1 0.990.91 0 0.980.990.980.97 1 1 1 1 1 0.98 1 0.990.990.91 0 0.910.990.980.970.99 1 1 1 1 1 0.970.980.970.91 0 0.910.990.980.970.99 1 1 1 1 1 1 0.950.890.71 0 0.910.99 1 0.970.99 1 1 1 1 1 1 1 0.890.71 0 0.710.970.99 1 0.98 1 1 1 1 1 1 1 1 0.71 0 0.710.890.98 1 1 1 1 1 1 1 1 1 1 1 0 0.710.890.950.970.98 1 1 1 1 1 1 1 1 1 1

1 ●

●

1 ●

−3

10 5

Performance Events

2 3

●

−3

−2

−1

Query index

Figure A.1: Pattern matching algorithm based on dynamic time warping. In this example the performance data and score are identical. a) Similarity matrix based on inter onset intervals (IOI) from recorded performance events and score events. b) Pitch class distance between performance events and score events. c) Final Cost Matrix is based on a one-matrix minus the IOI-Similarity Matrix multiplied with the Pitch-Similarity Matrix. The algorithm searches the path with the lowest cost though the matrix. The linear black line indicates a perfect match between performance and score. d) The step pattern of the dynamic time warping algorithm de nes rules during the search. This Rabiner-Juang step pattern allows to omit up to three events.

122

Chapter A. Additional material saxophone experiments

IOI â&#x2C6;&#x2019; Similarity

Pitch â&#x2C6;&#x2019; Similarity

10 5

0 0 0 0 0 0 0 0.020.02 0 0.050.110.29 1 0.290.11 0 0 0 0 0 0 0.020.030.04 0 0.110.29 1 0.290.110.05 0 0 0 0 0 0 0.030.050.07 0 0.29 1 0.290.110.050.03 0 0 0 0 0 0 0.050.110.17 0 1 0.290.110.050.030.02 0 0 0 0 0 0.050.110.290.55 1 0.290.110.050.03 0 0 0 0 0 0 0.020.110.29 1 0.550.290.110.050.03 0 0 0 0 0 0 0.020.030.29 1 0.290.170.110.050.03 0 0 0 0 0 0 0.020.030.05 1 0 0 0 0 0 0 0 0 0 0 0 0 0.030.050.110.29 0 0 0 0 0 0 0 0 0 0 0 0.030.050.110.290.110.050.030.020.020.01 0 0 0 0 0 0.020.050.110.29 1 0.050.030.020.020.01 0 0 0 0 0 0 0.030.110.29 1 0.290.030.020.010.01 0 0 0 0 0 0 0 0.050.29 1 0.290.110.020.010.01 0 0 0 0 0 0 0 0 0.11 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.290.290.110.050.030.010.01 0 0 0 0 0 0 0 0 0 1 0.110.050.030.020.01 0 0 0 0 0 0 0 0 0 0

Performance Events

10 5

Performance Events

1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1 0.050.050.050.110.29 1 0.29 1 0.050.110.110.29 1 0.050.050.05 0.110.110.110.29 1 0.29 1 0.290.110.290.29 1 0.290.110.110.11 0.290.290.29 1 0.290.110.290.110.29 1 1 0.290.110.290.290.29 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1 0.050.050.050.110.29 1 0.29 1 0.050.110.110.29 1 0.050.050.05 0.110.110.110.29 1 0.29 1 0.290.110.290.29 1 0.290.110.110.11 0.290.290.29 1 0.290.110.290.110.29 1 1 0.290.110.290.290.29 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1 0.050.050.050.110.29 1 0.29 1 0.050.110.110.29 1 0.050.050.05 0.110.110.110.29 1 0.29 1 0.290.110.290.29 1 0.290.110.110.11 0.290.290.29 1 0.290.110.290.110.29 1 1 0.290.110.290.290.29 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1 1 1 1 0.290.110.050.110.05 1 0.290.290.110.05 1 1 1

Score Events

Final Cost Matrix d)

Score Events

â&#x2014;?

â&#x2C6;&#x2019;1

â&#x2014;?

â&#x2C6;&#x2019;2

â&#x2014;?

1 Reference index

1 1 1 1 1 1 1 1 0.970.99 1 0.990.98 0 0.710.89 1 1 1 1 1 1 1 0.97 1 0.99 1 0.91 0 0.980.99 1 1 1 1 1 1 1 1 0.980.990.95 1 0 0.910.990.99 1 1 1 1 1 1 1 0.980.990.910.45 0 0.910.990.980.99 1 1 1 1 1 1 1 0.990.98 0 0.840.910.99 1 0.97 1 1 1 1 1 1 0.990.970.91 0 0.980.980.990.980.97 1 1 1 1 1 1 0.990.970.98 0 1 1 1 1 1 1 1 1 1 1 1 1 0.970.980.990.91 1 1 1 1 1 1 1 1 1 1 1 0.970.980.990.980.99 1 0.970.990.99 1 1 1 1 1 1 1 1 0.990.91 0 0.980.97 1 1 1 1 1 1 1 1 0.98 1 0.990.91 0 0.910.970.99 1 1 1 1 1 1 1 1 0.970.980.91 0 0.910.990.99 1 1 1 1 1 1 1 1 1 0.950.89 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0.890.710.710.970.99 1 1 1 1 1 1 1 1 1 1 1 0.71 0 0.890.98 1 1 1 1 1 1 1 1 1 1 1 1 0 0.890.950.970.980.99 1 1 1 1 1 1 1 1 1 1

1 â&#x2014;?

â&#x2014;?

1 â&#x2014;?

â&#x2C6;&#x2019;3

10 5

Performance Events

2 3

â&#x2014;?

â&#x2C6;&#x2019;3

â&#x2C6;&#x2019;2

â&#x2C6;&#x2019;1

Query index

Figure A.2: Pattern matching algorithm based on dynamic time warping. Example with di erences between performance data and score. a) Similarity matrix based on di erent inter onset intervals (IOI) from recorded performance events and expected score events. b) Pitch class distances between performance events and score events. c) Final Cost Matrix is based on a one-matrix minus the IOI-Similarity Matrix multiplied with the Pitch-Similarity Matrix. The black line shows the path found by the algorithm to match the performance to the score. The algorithm searches the path with the lowest cost trough the matrix. d) The step pattern of the dynamic time warping algorithm de nes the rules during the search. This Rabiner-Juang step pattern allows to omit up to three events.

Appendix B

Additional material clarinet experiments

123

124

Chapter B. Additional material clarinet experiments

    44                                          

Tempo: 120 bpm

         

       

        

      

Figure B.1: Melody designed for the warm-up task in the clarinet study (Chapter 4).

Slow Tempo

Fast Tempo

Low Dynamics

High Dynamics

Low Dynamics

High Dynamics

Low Register

High Register

Table B.1: Experimental design: 2 × 2 × 2 (tempo: slow fast, dynamics: piano forte, register: low high;), according score excerpts from Weber Clarinet Concerto, can be found in Appendix B.2 (A D) and B.3 (E H).

125

Chapter B. Additional material clarinet experiments

Allegro  3   4

 

118



 



  

              

Un poco ritenuto

   43

170



Cadenza

   43

159







 

   











                                                      Allegro Solo passinato

   43 

261

Figure B.2: Excerpts from the Clarinet Concerto No.1 in F minor (Op. 73) for clarinet in Bb, from C.M.v.Weber, in order of appearance in the piece. Bar numbers refer to the position in the rst movement.

126

Chapter B. Additional material clarinet experiments

         Adagio p

   73





                       



Adagio                  p



   3



   

     

poco cresc.

 

   

   

 

 

(Allegro)  2                                             4 p

                                       

(Allegro)                                                 

269





        

         

 





Figure B.3: Excerpts from the Clarinet Concerto No.1 in F minor (Op. 73) for clarinet in Bb, by C.M.v.Weber, in order of appearance in the piece. Bar numbers of E) and F) refer to bar numbers in the second movement, bar numbers of G) and H) refer the to third movement. Fingers + Tongue

Tongue only

 44                          1

Figure B.4: Stimuli used for the technical exercise task in the clarinet study (Chapter 4). 23-tone melody in B- at notation. Note numbers 1 8 require tonguing only. Note numbers 9 23 require sequential key-depression by left-hand ngers.

Chapter B. Additional material clarinet experiments

127

Figure B.5: Two participants closing all 6 sensor equipped tone holes of the sensor clarinet. Participant No. 2 (left) showed nger forces on an average level. In contrast participant No. 4 (right) showed the highest peak nger forces measured during the experiment (up to 12 N). Note the color of participant 4's nger tips. Figure B.6 and B.7 show the related force measurement.

1.0 â&#x2C6;&#x2019;1.0 0.0

0.6

0.2

0.8

1.4

0.0

0.6

0.0

Ring Sensor 1 (N)

0.2 0.6 1.0

Ring Sensor 2 (N)

3.0

Ring Sensor 3 (N)

1.5

Ring Sensor 4 (N)

0.0

Ring Sensor 5 (N)

1.5

Ring Sensor 6 (N)

Chapter B. Additional material clarinet experiments

0.5

Reed Signal

128

Time (s)

Figure B.6: Measured nger forces applied to the six ring shaped force sensors of the clarinet. Participant 2 performs the excerpt A (bar 118) from the Weber Clarinet Concerto No. 1 ( nd score in Appendix Figure B.2).

Chapter B. Additional material clarinet experiments

129

Figure B.7: Measured nger forces applied to the six ring shaped force sensors of the clarinet. Participant 4 performs the excerpt A (bar 118) from the Weber Clarinet Concerto No. 1 ( nd score in Appendix Figure B.2).

130

Chapter B. Additional material clarinet experiments

Figure B.8: Left-hand index nger force (in Newton) on clarinet, captured for participants 1 8 performing the Weber excerpt A. The plot shows two force curves (N), for the two captured trials under the highly expressive playing condition.

Chapter B. Additional material clarinet experiments

131

Figure B.9: Left-hand index nger force (in Newton) on clarinet, captured for participants 9 16 performing the Weber excerpt A. The plot shows two force curves (N), for the two captured trials under the highly expressive playing condition.

132

Chapter B. Additional material clarinet experiments

Figure B.10: Left-hand index nger force (in Newton) on clarinet, captured for participants 17 22 performing the Weber excerpt A. The plot shows two force curves (N), for the two captured trials under the highly expressive playing condition.

Chapter B. Additional material clarinet experiments

133

134

Chapter B. Additional material clarinet experiments