Meetingof
MoM_Digest_allPages.indd 1
Minds
the
4/20/11 3:03 PM
Meeting of the minds is an annual symposium at Carnegie Mellon University that gives students an opportunity to present their research and project work to a wide audience of faculty, fellow students, family members, industry representatives and the larger community. Students use posters, videos and other visual aides to present their work in a manner that can be easily understood by both experts and non experts.
Through this experience, students learn how to brindege the gap between conducting research and presenting it to a wider audience. A review committee consisting of industry experts and faculty members from other universities will review the presentations and choose the best projects and posters. Awards and certificates are presented to the winners.
MoM_Digest_allPages.indd 2
4/20/11 3:03 PM
Table of Contents
POSTER # TITLE
PAGE
Computer Science Posters
Q1 Integrating Natural Gestures into Touch Interfaces
1
Q2 iSTEP 2011 Montevideo, Uruguay
3
Q3 Leveraging Social Networking and Indoor Localization with Context-Aware Applications
5
Q4 Place Recognition for Indoor Blind Navigation
7
Q5 Realistic Face and Lip Expressions for a Bilingual Humanoid Robot
9
Q6 Spam Filtering by Na誰ve Bayes Classifier
11
Information Systems Posters
Q7 Assignment Deadline Tracker
13
Q8 ECShelf
15
Q9 Footprint Calculator
17
Q10 Improving the Deaf Dictionary
19
Q11 Intelligent Parking Guidance System
21
Q12 Kairion: Visualizing Rhetorical Source Data
23
Q13 Mezaniah
25
Q14 The Organ Donor Hub
27
Liberal Arts and Sciences Posters
Q15 Effects of Emotions on Mathematical Learning
29
Post-Graduate Posters
QG1 Advanced Water Treatment Method for Brackish Water Desalination
31
QG2 Arabic Natural Language Processing Research at CMUQ
33
QG3 Coverage in Visual Sensor Networks with Pan-Tilt-Zoom Cameras
35
QG4 Data Rate Based Aggregation to Improve Voice Capacity of Wireless LANS
37
QG5 MoBeNets: Mobile Behavior-based Networking
39
QG6 Interaction Engineering: Training of Wireless Interference
41
QG7 Interactional Disparities in English and Arabic Native Speakers with a Bilingual Robot Receptionist
43
QG8 LNG Pipe Mapping using Stereo Vision
45
QG9 Locality-Aware Reduce Task Scheduling for MapReduce in Cloud Computing
47
QG10 Micro-Benchmarks to Mitigate Performance Variations on the Cloud
49
QG11 On Packet Loss Rate and SINR Estimation in Wireless Local Area Networks
51
QG12 Video Aggregation: High-Quality Video Streaming in Multi Camera Surveillance Systems
53
MoM_Digest_allPages.indd 3
4/20/11 3:03 PM
MoM_Digest_allPages.indd 4
4/20/11 3:03 PM
MoM_Digest_allPages.indd 5
4/20/11 3:03 PM
Integrating Natural Gestures into Touch Interfaces
Authors Fahim Imaduddin Dalvi (CS 2014) Ameer Abdulsalam
Faculty Advisor Majd Sakr, Ph.D.
Category Computer Science
Abstract: This work aims to explore the role of Natural Gestures in daily interactions with computer systems, in particular, their use in the navigation of touchscreen interfaces. Gestures provide a way for users to navigate an interface through intuitive on-screen touch motions and are leading to a shift from traditional point and click interactions to a more natural and physical way of interaction. Given the increased popularity of public touchscreen kiosks in various settings such as airports, hospitals and company lobbies, we designed and built a test-bed platform for exploring touchscreen interface design for users of mixed lingual and cultural backgrounds. Inspired by the increasing prominence of gestures in commercial touchscreen devices, our aim was to explore the effects of language and culture on gestures, including the impact of various aspects such as screen size on the usability and practicality of these gestures. We implemented a few of these gestures into our interface, such as natural scrolling, which enables the user to flick their fingers across the screen in order to browse through a list of items. As part of future work, we seek to implement additional gestures, such as screen swiping, into the interface and to deploy this system on a kiosk at Carnegie Mellon Qatar’s campus for the purpose of collecting logs and running experiments. Through these experiments we seek to learn more about the interaction of users with the interface, their preferences and navigation performance, while considering the roles of language and culture in this region.
1
MoM_Digest_allPages.indd 6
4/20/11 3:03 PM
Integrating Natural Gestures in touch Interfaces
Fahim Imaduddin Dalvi {fid,msakr}@qatar.cmu.edu Ameer Abdulsalam ameer@cmu.edu Adviser : Dr. Majd F. Sakr Computer Science - Carnegie Mellon Qatar
The Problem Traditional Human-Computer interaction methods are not very intuitive and natural to the average user. Advent of touch screen technology has enabled more rapid and efficient interaction, especially on public information kiosks. This work aims to explore the design and implementation of effective touch screen user interfaces, in particular, touch gestures that cater to users of varying lingual and cultural backgrounds.
Implementation
User touches the screen
Finger is Lifted of Finger moves from one the screen. Calculate the Distance between point to another initial and final position
Using Distance and time, calculate the speed of the finger, and scroll accordingly. Therefore, If one flicks his finger faster, the list scrolls much faster.
Related Work The leading technologies today use multitouch on portable devices, which enables more complex gestures like pinch zoom. Single touch devices are primarily available in two types, surface capacitive, which is generally cheap and infrared touch screens, which are more accurate, but not very common.
Specifications and Constraints
Large screen size - Poses a challenge in implementing the gestures used in touch devices today, as it is difficult to perform the same gestures on a large screen. Single touch screen - Limits the possible gestures that can be implemented. Limited the interface to the English and Arabic knowledge, as these are the most popular languages in the region. Implement an up/down scrolling gesture.
Design
Adobe Flash and ActionScript 3. Object-oriented aspect of ActionScript 3 helped simplify the process a lot. The Gesture: Initial touch position on the screen and the direction of movement from initial position were recorded. Scrolling: Used timers to derive the speed at which the list should move.
MoM_Digest_allPages.indd 7
Results Successful implementation of scrolling gesture on large touch screen. Initial feedback is positive, gestures serves as a better alternative to the unnatural scroll bar interface. Due to the inaccuracy of the touch sensor on the touch screen, it results in some unpredictable behavior.
Future Work Implement right/left swiping gesture. Explore the potential of using multi-touch for large UIs as in information kiosks. Measure error and improve gesture recognition algorithm.
4/20/11 3:03 PM
iSTEP 2011 Montevideo, Uruguay Authors: Afnan Fahim (CS 2013) Asma Hamid (IS 2012) Elise Gonzales (CS 2012) Jonathan Beebe (CS 2012) Meghan Nahass (Policy and Mgt 2012) Hanae Timoulali (Information Systems Mgt Graduate Student) Roberto Lopez (Public Policy & Mgt Graduate Student) Vanessa Vazquez (Public Policy & Mgt Graduate Student)
Faculty Advisor: Yonina Cooper, Ph.D. M. Bernardine Dias, Ph.D.
Category: Computer Science
Abstract: The iSTEP (innovative Student Technology ExPerience) internship program offered by TechBridgeWorld at Carnegie Mellon University brings together students from the Pittsburgh and Doha campuses to conduct technology field research in developing communities. The 2011 iSTEP team is working with the Administración Nacional de Educación Pública and schools located in Montevideo, Uruguay on three projects: (1) a computer tool for teachers to customize literacy content; (2) an online computer tool for students to practice their literacy skills; and (3) a mobile phone game for motivating students to enhance their literacy skills. This poster outlines the team’s research about the relevant technologies that can be developed for increasing literacy skills, particularly in Uruguay. It also includes the research carried out about the country and its culture, which helped identify how the features of the technology could be utilized to make it relevant to the students in Uruguay. In the Spring 2011 semester, the team’s research included creating a country review investigating the history, culture, and technology infrastructure in Uruguay in order to be able to develop effective and easy to use tools for the school children in Uruguay. The iSTEP team also conducted interviews with locals to best understand the requirements and to make the tools culturally relevant. An intensive literature review of the different technological platforms available for portable use by students in Uruguay was carried out, which also involved testing different open-source projects. The technologies surveyed were based on both mobile and laptop devices, and for this purpose, appropriate protocols were developed to facilitate the dissemination of material onto such devices. The research allowed the team to identify the relevant technologies which that be utilized or developed in order to create culturally relevant tools that would contribute to increasing the literacy levels of school children in Uruguay.
3
MoM_Digest_allPages.indd 8
4/20/11 3:03 PM
iSTEP 2011 Montevideo, Uruguay Background Research and Technology Evaluation
Afnan Fahim - afahim@qatar.cmu.edu | Asma Hamid - amhamid@qatar.cmu.edu Yonina Cooper Ph.D | Computer Science
Motivation
To assist the partner in Uruguay in developing technology solutions to enhance English literacy skills.
Goal
• In order to understand what factors would determine the success of literacy tools in target community in Uruguay, research needs to be carried out on its culture, traditions and educational system of the target community. • Appropriate technology must be used to develop these literacy tools, and to ensure that, different technological platforms need to be assessed to find the best possible solution.
Approach
1. Conducted interviews with teachers to understand Uruguay’s educational system. 2. Carried out literature review on the country, its educational system and development indicators. 3. Conducted literature review of technology infrastructure in Uruguay. 4. Identified possible technologies to develop literacy tools, based on the literature review outcomes. 5. Tested technologies and open-source alternatives to decide which ones best meet projects’ requirements.
Partners
Our partner for iSTEP 2011 is the ANEP, National Administration of Public Education in Montevideo, Uruguay who identified three projects.
Projects
OLPC/Computer Tool (OCT): Culturally relevant English literacy tool accessible on the OLPC and on regular laptop and desktop computers. Content Authoring Tool (CAT): Computer tool for teachers and English literacy experts in Uruguay to customize literacy content. Mobile Phone Tool (MPT): Engaging mobile phone-based tool for students to practice English literacy skills.
Future work
• Collaborating with partners to further develop and customize tools • In-situ needs assessment and technology evaluation using user surveys and on-site interviews.
Results
• Uruguayan culture consists of a great deal of fascination with soccer and the country’s historical roots. This is now being integrated into the content of the MPT tool and OCT. • J2ME platform is the most used by target audience, and is thus being considered as viable platform for MPT project. • In-situ needs assessment plans have been created based on outcomes of literature reviews and questions. • For identified platforms, development plans have been created for the three literacy tools. • Media plan has also been developed to inform the community about development of the literacy tools.
MoM_Digest_allPages.indd 9
Acknowledgements Silvia Pessoa, Bernardine Dias, Ermine Teves, Sarah Belousov
4/20/11 3:03 PM
Leveraging Social Networking and Indoor Localization with Context-Aware Applications Author Dania Abed Rabbou (CS 2012)
Faculty Advisor Khaled Harras, Ph.D.
Abstract: In the recent years, social networking has gained unprecedented popularity among Internet users. Online social network applications, such as Facebook, have hundreds of millions of users actively connecting through their mobile devices. However, the tight coupling between mobile users and their social relationships remains largely unexplored. Our research is aimed at using users’ profile information, given users’ consent, in conjunction with indoor localization to create a multi-purpose engine. Once a user is logged on to Facebook, the engine will request access to profile information. An embedded localization engine will subsequently report the user’s location. Having fetched the data and localized the user, the engine will perform rigorous profile matching against other Facebook users in the vicinity. Finally, it will report back all the matches and their locations in the given area. Such an engine has a wide variety of applications; the immediate one being a friend finder. The application Friend Finder will use our engine to suggest friendship to Facebook users based on mutual commonality and location. Another such application is in commercialization, where entrepreneurs can merchandize products to customers in a given area who match a certain criteria.
5
MoM_Digest_allPages.indd 10
4/20/11 3:03 PM
MoM_Digest_allPages.indd 11
4/20/11 3:03 PM
Place Recognition for Indoor Blind Navigation Authors Samreen Anjum (CS 2011)
Advisors Brett Browning, Ph.D. M. Bernardine Dias, Ph.D.
Category Computer Science
Abstract In Computer Vision, place recognition addresses the problem of identifying places based on their appearance. It has various applications including indoor navigation for the visually impaired. Navigating inside buildings is often a straightforward task for sighted people but can be challenging for the visually impaired. This often results in a lack of independence and privacy. There have been several approaches taken to develop indoor blind navigation systems, such as Barcode detectors, RFID information grids, Wifi and Sensors. However, in most of these approaches, there exist limitations such as aiming the camera, scalability, and lack of orientation information. In this work, we aim to develop a portable system with place recognition and navigation algorithms to enable independent navigation for the visually impaired in GPS denied indoor environments. Our approach focuses on separating the problem into two parts: mapping and localization. The mapping or training phase is an offline stage where a map of the indoor environment is built by a sighted mapper and stored for use in the localization phase. In operation, the images from portable cameras will be used to register, or localize, a visually impaired user against the map. After localization, the system will find the shortest path to the destination, and guide the user to the desired location. A successful navigation must first ensure that the user has been localized effectively. Hence, place recognition forms a vital component in indoor blind navigation and is the primary focus of this work. We investigate four different methods and their various combinations for user localization. The system is evaluated for image retrieval and image labeling inside the Carnegie Mellon University Qatar building, and the results are presented in this poster.
7
MoM_Digest_allPages.indd 12
4/20/11 3:03 PM
MoM_Digest_allPages.indd 13
4/20/11 3:03 PM
Realistic Face and Lip Expressions for a Bilingual Humanoid Robot Authors Amna Al-Zeyara (CS 2014) Nawal Behih (CS 2014)
Faculty Advisor Majd F. Sakr, Ph.D.
Category Computer Science
Abstract Hala is a bilingual (Arabic and English) robot receptionist located at Carnegie Mellon in Qatar. Users type to her, and she replies in speech along with realistic facial expressions. However there are two existing problems with the current model. Her Arabic lip-movements are not realistic because they were adopted from English utterances. Secondly, her animation engine does not adhere to existing research on face dynamics, which makes it difficult to leverage/create natural and interesting facial expressions. In this work we address these two limitations. The sounds humans make while talking are symbolically captured as “phonemes.” The corresponding shapes of the lips, for these sounds (i.e. phonemes), are called “visemes.” We used existing research to develop Arabic visemes that accurately capture the Arabic language and its pronunciations. Hala can utilize English visemes, but now also additional Arabic visemes for visually realistic lip-movement and syncing. Hala’s current facial animations are not rich. We used Paul Ekman’s seminal work on Facial Action Coding System (FACS) to demarcate her face into basic muscle-primitives. These muscle-primitives were used to compose complex, yet natural, facial expressions. Additionally, FACS categorizes all possible human facial emotions and we implemented a subset of these. We also identified existing limitations with the model that prevented us from creating certain facial expressions.
9
MoM_Digest_allPages.indd 14
4/20/11 3:03 PM
MoM_Digest_allPages.indd 15
4/20/11 3:03 PM
Spam Filtering by Na誰ve Bayes Classifier Author Manoj Dareddy (CS 2013)
Faculty Advisor Behrang Mohit, Ph.D.
Category Computer Science
Abstract Automatic filtering of spam emails is a problem which saves a lot of resources, and researchers in statistical natural language processing have been studying this problem for the past decade. This work focuses on developing a machine learning classifier for the detection of spam messages. The classifier is built upon a simple statistical model called Naive Bayes. Naive Bayes views text as a bag of words. Each email class (spam vs. legitimate) is represented by a different set of terms along with their frequencies. We first use a set of spam and legitimate emails to learn the probabilities and train the model. The learned model is then applied and evaluated against an unseen test data. The classifier and its standard evaluation metrics (precision and recall) were implemented in Python programming language. Furthermore, the classifier was trained and evaluated under different data and feature configurations. Experiments show a steady improvement of performance when training data is increased and the feature set is chosen selectively.
11
MoM_Digest_allPages.indd 16
4/20/11 3:03 PM
MoM_Digest_allPages.indd 17
4/20/11 3:03 PM
Assignment Deadline Tracker Authors Al-Amri, Amal (IS 2012) Al-Sooj, Ghada (IS 2012) Masood, Aamir (IS 2012) Reeves, Allan (IS 2012)
Faculty Advisor Ian Lacey, Ph.D.
Category Information Systems
Abstract Assignment Deadlines Tracker (ADT) is a collaborative calendar management software that exploits the power of the DRY and SSOT principles to add a synergistic value to the routine practice of creating, managing and following academic deadlines, which happens between professors and students on college campuses all over the world. In ADT, professors can create assignment deadlines for the courses they teach on one central calendar. These assignment deadlines are then automatically added to the individual calendars of all the students who are registered for that particular course. In this way the student community can have the information they need to plan their semesters well in advance without the individual effort on the part of each student in reading the syllabuses and manually entering all this information into his/her calendar over and over again. The second advantage this system yields is that the students can see the assignment/homework deadlines alongside the schedule for their personal activities and on-campus events which enables them to plan even more effectively. The third advantage of this software is the ability to export data to other applications like iCal, Outlook and GoogleCalendar which means that the software allows the integration of information across various software platforms. Last, and a very significant advantage of the system, is that whenever a professor makes a change in assignment-related details, all the students are automatically notified about the updates and the information is automatically updated in every student’s personal calendar without any human effort.
13
MoM_Digest_allPages.indd 18
4/20/11 3:03 PM
MoM_Digest_allPages.indd 19
4/20/11 3:03 PM
EC Shelf Project Authors Alkaabi, Rashid (IS 2012) Datta, Olympia (IS 2011) Hamid, Asma (IS 2012) Kelzieh, Abdulmunim (IS 2012)
Faculty Advisor Divakaran Liginlal, Ph.D.
Category Information Systems
Abstract The poster presents the Bookshelf project, which aims to facilitate and organize the process of book donation and the communication between the donors and charity organizations. The system aims to improve the current process of book donations, which is manual and prone to errors, besides ensuring wider dissemination of information to charity organizations. The project targets education city students and charity organizations in Qatar. We thought about this project after realizing the problems that the current process has. The current process involves distributing posters in the students’ dorms, asking them to leave their books in specific locations. There are several problems with the current process. First, it does not allow charity organizations to reach all targeted students/book owners due to the poor advertising. Second, it is a very slow process and takes a long time to categorize the books. Third, the way/place where books are collected may result in the books being destroyed or damaged. Interviews with potential stakeholders of the Bookshelf system were conducted to discuss the need and possibility of implementing the project. Results of the interviews with stakeholders such as Bijan Esfahani, the Director of our Library, and charity organization representatives will be presented in the poster. The Bookshelf project will save thousands of books that are wasted at the end of each semester. We believe that there are students who would like to donate their books and this system will facilitate the process for them. On the other hand, the system will help charity organizations reach more students as the application is linked to the social networking sites that are used by most of them. This work provides a deeper understanding of the use of Information Systems to promote social causes and social entrepreneurship.
15
MoM_Digest_allPages.indd 20
4/20/11 3:03 PM
MoM_Digest_allPages.indd 21
4/20/11 3:03 PM
Footprint Calculator Authors Samee Amin (IS 2011) Leah Hayes (IS 2011) Rob Houle (IS 2011) Robert Starkey (IS 2011) Libby Adams (IS 2011)
Faculty Advisor Randy Weinberg, Ph.D.
Category Information Systems
The Client Mr. John Leigh, the Recycling & Waste Programs Manager at the Dartmouth Hitchcock Medical Center, and Mr. Randy Pratt, the lead designer, have teamed up to create an ecological footprint calculator targeted at health institutions. The calculator covers all categories that may affect the estimated footprint from Products to Transportation. Dartmouth-Hitchcock Medical Center (DHMC) is New Hampshire’s only academic medical center.
The Problem The client was struggling with effectively sharing the calculator across the proper distribution channels. Originally, the entire project was encapsulated into a single Excel spreadsheet with many tabs. This proved to be an ineffective way to properly express all of the information the team wanted to share.
The Solution Our team delivered an online solution that allows the client to host the calculator on the web while providing a more appropriate avenue for sharing relevant information. By creating a Registration form and comments section, the clients are now able to begin seeing the global reach their calculator has. The site was created and hosted with Google sites so that when the transition to the DHMC team occurs, the ability to update the site will remain simple. The calculator spreadsheet also received several cosmetic and functional changes to improve usability. In addition to working with the DHMC team, our team also adapted the calculator for use with an academic institution. By adapting those categories relevant to a University campus (i.e. Agriculture and Energy), we hope to share our tool with the greater academic community and help campuses control their ecological footprint.
17
MoM_Digest_allPages.indd 22
4/20/11 3:03 PM
MoM_Digest_allPages.indd 23
4/20/11 3:03 PM
Improving the Deaf Dictionary Author Maha Al-Moghany (CS 2012)
Faculty Advisor Selma Limam Mansar, Ph.D.
Category Computer Systems
Abstract According to the Human Rights, the rights of people with deaf disabilities are inadequately protected. Deaf people use sign language to interact with each other and participate in modern-day society. However, the sign language is not unified in the Arab world, hence each country has its own sign language (i.e. its own deaf dictionary). Deaf people have limited interactions with their society due to the poor ways of communication between deaf and hearing individuals. It’s even more difficult to communicate between deaf people from different countries. Today the Arab world is developing a unified sign language that can be used globally within the Arab world. Through extensive interactions with the Qatari Cultural Center for Deaf, a cultural center for deaf people in Qatar, we found that adding a word into a deaf dictionary is always a non-trivial process that takes months if not years, in both the unified and non-unified dictionaries. Keeping the deaf dictionary updated with new terminologies that are used in our daily life is critical to allow deaf people to be full participants in modern-day society. This study addresses in depth the process of adding a word into the unified sign language dictionary. The outcome of this study will be detailed design of a social networking site. The site will help expertise to add words efficiently and in a timely manner as well as enhancing deaf individuals’ interactions and communications with each other, and be efficiently updated with the recent signs that have been added into the dictionary.
19
MoM_Digest_allPages.indd 24
4/20/11 3:03 PM
MoM_Digest_allPages.indd 25
4/20/11 3:03 PM
Intelligent Parking Guidance System Author Jimmy Musili (BA 2013) Faculty Advisor Daniel C. Phelps, Ph.D.
Category Information Systems
Abstract Communicating information about a parking spot to a car driver while he/she is a mile away from the parking area is a fundamental element of parking management around the world’s cities. Parking problems are the major contributor to drivers’ frustration, stress, road congestion, traffic, pollution and decrease of revenues for businesses. In tackling parking problems, a better intelligence device seems to be the solution to parking problems in our urban cities. The proposed idea makes use of sensys sensors, wireless communication and Google maps. The technology employs magneto-resistive sensors implanted in parking spots to detect the presence and movement of cars and thus transmit (YES) vehicle detected and (NO) empty spot to the access point. The access point then reports key parking information to the main parking station which integrates the data with Google maps and thus provides car drivers/mobile users/PDAs with real time information of available/taken spots. The core objective of this paper is to demonstrate the capability of sensors and wireless technology in combating parking problems by providing enough information to vehicle drivers and reducing their search time for parking.
21
MoM_Digest_allPages.indd 26
4/20/11 3:03 PM
MoM_Digest_allPages.indd 27
4/20/11 3:03 PM
Kairion: Visualizing Rhetorical Metadata Author Zaid Haque (IS 2012)
Faculty Advosor Andreas Karatsolis, Ph.D.
Category Information Systems
Abstract Current software that aims to support academic research (EndNote, Papers) only consider the descriptive metadata of sources (author name, publication date, etc) as the information that would interest researchers. However, extensive research in citation practices (Cronin 1984, Borgman 2007, Olson 2008) proves that researchers choose papers to cite based on more rhetorical information. Further research (Karatsolis 2005) led to the creation of a rhetorical model that suggests the information that researchers need when citing a source in their paper. Kairion (Greek, meaning ‘timely’) was developed as an electronic method to track the subjective nature of such rhetorical information. The idea is to visualize this rhetorical information of every source that the researcher has looked over while doing a certain project as a means of managing their library of research papers. In this system, the researcher, when reading a paper, will be asked a set of questions. These questions are aimed at providing a sense of what subjective information the researcher wants to remember about the paper in question. An example of a possible question would be “Your current work extends some of the work presented in this source” because this type of information is critical in determining how you would use this source in your own paper. When the information collected from the user is sent to the database containing the information from all the sources, it is visualized in a 3D environment as a means of keeping track of all the papers that the researcher read. The idea of Kairion was a product of years of research by Professor Karatsolis. The data involved in the production of the prototype is from PubMed. As the UI/Graphic Designer, I have been working on the system’s interface.
23
MoM_Digest_allPages.indd 28
4/20/11 3:03 PM
MoM_Digest_allPages.indd 29
4/20/11 3:03 PM
Mezaniah: A Budget tracking tool for CMUQ Student Organizations Authors Yazan Abu Hijleh (IS 2012) Zaid Haque (IS 2012) John-Paul Menniti (IS 2012)
Faculty Advisor Ian Lacey, Ph.D.
Category Information Systems
Abstract Student-run organizations often face problems when planning their expenses, and these problems prevent clubs from organizing events efficiently. There are three major problems with the current system at CMUQ. The first problem is overspending, which occurs when clubs go over budget. If clubs spend over their budget, they do not receive any money from Student Majlis and have to resort to spending their own money to pay for items and services. This is a large disincentive to organizing events and causes stress and frustration for clubs and their advisors. The second problem is underspending. For fear of spending too much, club leaders may elect to spend very little throughout the year. This causes the clubs to be absent from the student life scene at CMUQ. The third problem is accountability. It is hard to find out how much a club has left to spend in the current system, because Student Affairs Advisors must manually go through each club’s transactions, adding them up to calculate the balance. Student Majlis has to repeat this effort or communicate with advisors when figuring out how clubs are performing. The problems outlined previously affect all students in CMUQ, as even those who do not belong to clubs are missing out on everything clubs could be offering. Through research and implementation as part of 67373, our project team developed Mezaniah. The idea for the project came to us as we are all affiliated with clubs at some level and know their budgeting shortcomings. Mezaniah supports CMUQ Student Organizations and manages their budgets.
25
MoM_Digest_allPages.indd 30
4/20/11 3:03 PM
MoM_Digest_allPages.indd 31
4/20/11 3:03 PM
The Organ Donor Hub Authors Mohamed Al Haddad (IS 2012) Ghanim Al Sulati (IS 2012) Fatima Mujahid (IS 2012) Rana Khalil (IS 2012)
Faculty Advisor Divakaran Liginlal, Ph.D.
Category Information Systems
Abstract Diabetes, obesity and genetic diseases have a great impact on Qatar’s populace, consequently increasing the annual need for kidney transplants. Since 1986, Qatar has initiated a Kidney Transplant Program to consolidate the donation of kidneys. The lack of kidney donors in Qatar has been a great challenge for the program resulting in 20% deaths at end stage kidney disease. The State of Qatar is taking steps to make this program the hub for renal transplant (kidney transplant) in the region and to serve its mission for a better health care for the 2030 national vision. As part of the community, our team decided to take action and assist this development and build a system to help clinics and hospitals match donors with patients for possible successful renal transplants. Our Kidney Donation system contains a Management System for both patients and donors, and a website to raise awareness in the community. The System then integrates the Donor Management System and the Patient Management System and through an interface matches patients against donors (and donors against patients) and generates reports for the results. The goal is to make the matching process for both patients and donors reliable, more time efficient and accurate. This system will impact the clinics and hospitals in Qatar as until now there is currently a high reliance on human effort and paperwork to perform the match. An automated system will help hospitals keep track of new patients/donors, reduce error and produce the matching output in a clear and visually comprehensible format. Doctors and labs will be able to track critical cases, keep record of all information activities and in the process, save even more lives.
27
MoM_Digest_allPages.indd 32
4/20/11 3:03 PM
MoM_Digest_allPages.indd 33
4/20/11 3:03 PM
Effects of Emotion on Mathematical Learning Author Lulwa Ahmed El-Matbouly (CS 2013)
Faculty Advisor Angela Brunstein, Ph.D.
Category Liberal Arts & Sciences
Abstract This research investigated the impact of emotion on learning and applying mathematical procedures. The purpose of the study is to evaluate the effects of emotional arousal on the ability to learn and remember factual information and mathematical procedures. Participants attended three sessions, one hour per each. For the first session, the students mainly learned a relaxation technique and got familiar with the experimental setting and procedure. In the second session, students watched 2 videos, the first one presents a cover story similar to the ones used in mathematical text books. The second video presents a procedure for solving integration problems. In the third session, students applied what they had learned to a new problem. These three sessions help in distinguishing the main factors that the students get affect by. The main goal of the session is to test the performance of the students and how it varies depending on their skin resistances and emotions toward math. In addition, the relaxation technique gives us standard base to compare the students. Since there are two different stories, we will examine the impact of these stories. We assume that the students either like the stories then perform well in solving the math problem, or they don’t like the story and perform weakly in solving the math. Mainly, the stories control how the students get involved in understanding the concepts and make solving the math problem more interesting and easy to remember. From the overall sessions, we noticed that the students become more relaxed and calm after the relaxation session. Moreover, the stories are the main factor that controls the performance of the students.
29
MoM_Digest_allPages.indd 34
4/20/11 3:03 PM
MoM_Digest_allPages.indd 35
4/20/11 3:03 PM
Advanced Water Treatment Method for Brackish Water Desalination Author Lichen Han Khrishnapuram Karthikeyan
Faculty Advisor Khrishnapuram Karthikeyan, Ph.D.
Abstract We aim to develop an energy-efficient, low-cost desalination technology for creating new, affordable water sources from brackish waters. Since Qatar has extremely limited rechargeable water sources, technologies facilitating safe use of impaired and unconventional water sources are needed. Capacitive deionization (CDI) technology can meet the unique, logistical and economic needs for inland desalination. CDI technology, unlike reverse osmosis and nanofiltration, operates under ambient pressure and can be sustained with renewable energy sources. In this process, ions move to the electrode surface and build up an electrical double layer (EDL) when an external potential is applied. Metal oxide nano-particle coatings are widely used in supercapacitors to increase the capacitance. We are employing fourth-generation CDI technology (i.e., alumina, silica nano-particle coated porous carbon) for brackish water treatment. The presence of oxides together with the functional groups on the carbon surface enhances the ion removal process. Our system can remove 60% NaCl under optimal experimental conditions with high reversibility and no obvious loss of capacity in longterm operation. During regeneration, up to 50% of the charge can be harvested. Since EDL charging is a fast process (a few seconds), desalination rate is limited by the diffusion of ions within the porous electrodes (> 1 hour). Tests are underway for various electrolytes (1:2, 2:1, 2:2) and brackish water samples to cover a wide gradient in salt concentration and ionic composition. This project will lead to the development of low-cost inland desalination systems and can be expected to boost Qatar’s scientific profile in the global and regional water industry.
31
MoM_Digest_allPages.indd 36
4/20/11 3:03 PM
MoM_Digest_allPages.indd 37
4/20/11 3:03 PM
Arabic Natural Language Processing Research at CMUQ Author Behrang Mohit Rishav Bhowmick
Faculty Advisor Behrang Mohit, Ph.D. Kemal Oflazer, Ph.D.
Abstract This poster presents an overview of our ongoing research of using statistical natural language processing for: (1) Identifying named entities (NEs) in Arabic Wikipedia articles. (2) Unsupervised clustering of named entities based on meta-data extracted from Wikipedia articles. (3) Automatic translation of Wikipedia articles from English to Arabic. Working on Wikipedia text has the challenge of lacking in-domain labeled data for most of the topics and articles. For statistical modeling of these tasks, we benefit from semi-supervised learning and domain adaptation techniques. With these techniques we enrich the statistical models, by exploiting knowledge from out-of-domain and unlabeled data.
33
MoM_Digest_allPages.indd 38
4/20/11 3:03 PM
MoM_Digest_allPages.indd 39
4/20/11 3:03 PM
Coverage in Visual Sensor Networkswith Pan-Tilt-Zoom Cameras Authors Vikram P. Munishwar (Researcher) Vinay Kolar, Ph.D. Nael B. Abu-Ghazaleh, Ph.D.
Category Computer Science
Abstract: Visual sensor networks (VSNs) are networks made up of a possibly large number of cameras that together monitor an area of interest. VSNs are finding increasing applications in surveillance, traffic, border control and other monitoring applications. We consider the coverage problem for Pan-Tilt-Zoom camera networks, with the goal of maximizing the number of targets monitored by the cameras. The degrees of freedom in the selection of the Field of View in PTZ cameras make the problem difficult. First, we formulate and solve the single camera problem of identifying the FoVs representing the maximal subsets of targets (maxFoV). We show that maxFoV, under general assumptions, is NP-hard. However, the geometric nature of the problem makes the problem polynomial, which we demonstrate by constructing an optimal algorithm to solve the problem with a worst case complexity of O(n^3). This is a surprising result as maximal clique problems are NP-complete for other geometric graphs such as unit disc graphs. We prove the correctness of the algorithm and experimentally study its performance and run time. We integrate the algorithm with multi-camera coverage maximization algorithms, and study its impact on the quality of coverage and run time. We also compare the gain in coverage obtained from PTZ cameras relative to fixed FoV cameras for a number of scenarios. Finally, we implement and evaluate the algorithms in a small testbed to demonstrate that the approach is practical.
35
MoM_Digest_allPages.indd 40
4/20/11 3:03 PM
MoM_Digest_allPages.indd 41
4/20/11 3:03 PM
Data Rate Based Aggregation to Improve Voice Capacity of Wireless LANS Author Adnan Majeed
Faculty Advisor Nael Abu-Glazaleh, Ph.D.
Abstract Researchers predict cellular companies will loose $5 billion in revenue by 2015 due to a significant increase in VOIP over Wifi (VoWifi) usage. Poor VoWifi performance can hinder this growth. Wifi Access Points (APs) support a very small number of VOIP calls due to the MAC overhead of sending small sized VOIP packets frequently. Researchers show that using packet aggregation can increase this number. Previous works implement aggregation in one of two ways, each with its own drawback. Either (1) packets for a particular next hop are aggregated and unicast to that node or (2) as many packets as possible are aggregated and broadcast to all nodes. The former mechanism has limited gain in Wireless LANs (WLANs) since the number of packets to aggregate is very small. The latter mechanism, which we refer to as basic aggregation, requires that all nodes should be able to receive the broadcast packet correctly. Therefore, the packet has to be broadcast at the data rate of the link with the lowest data rate. This severely limits the gain from aggregation and in certain cases leads to a loss in performance. This poster proposes a Data Rate based Aggregation (DRA) protocol that does not suffer from either of these problems. DRA divides packets in the AP queue into groups based on the data rate they are to be transmitted at. The algorithm aggregates packets in the same group and broadcasts the aggregated packet at the data rate of that group. In the best case DRA can lead to a 120% to 250% increase in the number of calls supported by a single 802.11b AP; in random cases we observe a 45% increase in the number of calls supported by a single 802.11b AP.
37
MoM_Digest_allPages.indd 42
4/20/11 3:03 PM
MoM_Digest_allPages.indd 43
4/20/11 3:03 PM
Interaction Engineering: Taming of Wireless Interference Author Saquib Razak, Ph.D. Vinay Kolar, Ph.D.
Faculty Advisor Nael Abu-Ghazaleh, Ph.D.
Abstract Carrier Sense Multiple Access (CSMA) protocols are unable to effectively arbitrate the medium in multi-hop wireless networks; problems such as hidden and exposed terminals occur frequently leading to collisions, poor performance and unfairness. CSMA networks can be optimized by careful tuning of transceiver parameters, such as transmission power and carrier sensing threshold, to maximize transmission concurrency while minimizing collisions. Existing approaches optimize these parameters by considering only some aspects of CSMA operation (e.g., considering only PHY parameters such as SINR), thus leading to suboptimal solutions. We present a new approach that leverages recent insights into the behavior of CSMA networks. Specifically, this recent work identifies that links interfere only in a few discrete interaction modes at the MAC-layer. Each interaction mode determines how the interacting links interfere with each other and leads to different performance and fairness behavior. The proposed methodology controls the transceiver parameters to convert the destructive interactions into constructive ones; we call this approach Interaction Engineering. The global optimization problem is computationally infeasible and requires a central solution. Therefore, we first formulate an interaction engineering model that computes the parameters based on one-to-one interaction between the links. Second, we extend the model into a distributed interaction-aware MAC protocol (I-MAC). Testbed and simulation results show that collisions and retransmissions are almost completely eliminated, leading to large improvements in throughput and delay.
39
MoM_Digest_allPages.indd 44
4/20/11 3:03 PM
MoM_Digest_allPages.indd 45
4/20/11 3:04 PM
Interactional Disparities in English and Arabic Native Speakers with a Bilingual Robot Receptionist Authors Imran Fanaswala
Faculty Advisors Brett Browning, Ph.D. Majd Sakr, Ph.D.
Abstract Hala is a permanently deployed, bilingual (Arabic and English) and culturally sensitive robot receptionist located in Carnegie Mellon University in Qatar. She caters to a multi-cultural demographic and her personality is that of a young, unmarried, Arab female receptionist. Hala serves as a research testbed for studying the influence of socio-cultural norms and the nature of interactions during human-robot interaction within a multicultural, yet primarily ethnic Arab, setting. Current work studies how the Arab culture reacts to a roboceptionist, and conversely, if her Arabic personality elicits di fferent responses. Now, HRI studies in a Middle Eastern environment are subject to nuances and subtleties. This study explores the nature of interactions, in an uncontrolled environment, between Hala and users of varied native tongues. We correlate an interlocutor’s native language with their propensity for accepting an invite and the duration of the ensuing conversation. Subsequently, we present results that demonstrate significant disparity in interactional patterns between English and Arabic speakers. Since most of Hala’s users are English speaking but of Arabic ethnicity, we also assess the importance of a transliterated Arabic input mode. After 6 months of continuously running experiments, our results indicate that Arabic native speakers were twice as likely to accept an invite and tended to converse for 25% longer compared to their English counterparts. We also realised that approximately 70% of Arabic input was transliterated, thus signifying its importance as an input technique for Arabic speakers. Finally, this study of native language is a segue towards the inquiry of culture as a factor during human robot interaction in a multi-cultural setting.
41
MoM_Digest_allPages.indd 46
4/20/11 3:04 PM
MoM_Digest_allPages.indd 47
4/20/11 3:04 PM
LNG Pipe Mapping using Stereo Vision Author Peter Hansen Peter Rander Hatem Alismail Samitha Ekanyake Mohamed Mustafa
Faculty Advisor Brett Browning, Ph.D.
Abstract: Liquified Natural Gas (LNG) production is a key industry in Qatar. Pipe corrosion can be a critical safety hazard in this industry, and regular inspection is therefore necessary to prevent failures. In current industry practice, corrosion can be monitored using periodic measurements of pipe wall thickness. However, sensors such as magnetic flux leakage (MFL) must often be used external to the pipe and may not provide full coverage. To overcome these limitations we are developing a visual perception system for automated corrosion detection in LNG pipes. Our current system uses a stereo camera mounted on a pipe crawling robot. The imagery from the camera is used to generate high resolution appearance maps of the internal pipe surface which include both appearance and sub-millimeter 3D structural data. Either using these maps directly, or by registering their change over time, suitable algorithms will be developed for automatically detecting areas of corrosion. A core requirement for building these maps is to find highly accurate position estimates of the robot. We use the images to estimate robot position within approximately 0.1 percent error in distance travelled.
43
MoM_Digest_allPages.indd 48
4/20/11 3:04 PM
MoM_Digest_allPages.indd 49
4/20/11 3:04 PM
Locality-Aware Reduce Task Scheduling for MapReduce in Cloud Computing Authors Mohammad Hammoud (Postdoctoral Research Associate, Cloud Computing Lab) and Majd Sakr
Faculty Advisor Majd Sakr, Ph.D.
Category Computer Science
Abstract Due to its extraordinary features in minimalism, fault tolerance, and scalability, MapReduce is by far the most successful realization of data intensive cloud computing platform. In the MapReduce framework, any application is specified by jobs. The input to a MapReduce job is divided into fixed-size pieces denoted as splits. To process splits concurrently and independently, each split is assigned a map task by a master node (i.e., oneto-one assignment). Being aware of the network locations of splits, the master node attempts to schedule map tasks in the vicinity of splits in order to diminish data transfer in an environment typified by scarcity and preciousness in network bandwidth. In contrast, the master node disregards data locality when scheduling reduce tasks. This is because the input to a single reducer is normally the output from many mappers (i.e., many-to-one assignment). In fact, the reduce task scheduling determines the pattern of the communication traffic in the cloud and, as such, renders a vital part that affects the completion time of the whole job. This work investigates the problem of locality unawareness in reduce task scheduling and promotes LocalityAware Reduce Task Scheduler (LARTS), a practical strategy for providing high-performance MapReducebased systems. LARTS schedules reducers at the center of gravity from all the feeding mappers in order to minimize the network traffic induced by the shuffling of data and maximize the overall throughput of the system. To select the mass center for a reducer, LARTS requires acquaintance of twofold: (1) the network locations of all the feeding mappers and (2) the distances between nodes in the cluster. While the master node already maintains the latter, the former can be easily established via involving the heartbeat mechanism of the MapReduce model. LARTS is currently under development over the native Hadoop-0.20.2.
45
MoM_Digest_allPages.indd 50
4/20/11 3:04 PM
MoM_Digest_allPages.indd 51
4/20/11 3:04 PM
Micro-Benchmarks to Mitigate Performance Variations on the Cloud Author Suhail Rehman
Faculty Advisor Majd Sakr, Ph.D.
Abstract Cloud computing offers computing resources to users on an on-demand, pay-as-you-go basis. Public clouds such as Amazon’s Elastic Compute Cloud (EC2) allow their clients to quickly provision Virtual Machines (VMs) with the required software/libraries, and rapidly change or scale compute resources according to demand. A primary advantage of cloud computing is the reduction of upfront investment, thereby allowing clients to be billed for the resources that they use for a particular amount of time. Cloud providers provision virtualized resources (VMs with virtual CPUs, Memory and Disk Space) to clients attempting to carefully trade-off profit with quality of service. Ultimately, the underlying physical resources that are mapped to virtualized resources are likely to be shared among multiple clients. Sharing of physical resources improves their overall utilization, but introduces variations in load, which affects performance, particularly during peak loads. Cloud providers currently do not guarantee strict performance isolation; the load on a physical resource by one client can potentially affect the availability and performance of the same resource when being shared by another client. Only the cloud provider can monitor and control how the resources have been provisioned among clients; the client can simply configure and request virtualized resources. As a result, several studies have shown that there is a high variability in the performance of applications deployed in such cloud environments. Information regarding the load on the underlying physical resources would be valuable in order to manage the potential performance variations when assigning tasks on virtualized resources. In our work, we utilize microbenchmarks to assess available CPU power, memory and disk bandwidth, as well as point-to-point network bandwidths. Careful analysis of the performance of the micro-benchmarks yields a better understanding of (a) the relative mapping of the virtualized resources to the physical resources, and (b) the relative load of the underlying physical resources. We use this information to make changes in the mapping and scheduling of a distributed application (such as a Hadoop MapReduce application) to limit performance variations during peak loads.
47
MoM_Digest_allPages.indd 52
4/20/11 3:04 PM
MoM_Digest_allPages.indd 53
4/20/11 3:04 PM
MoBeNets: Mobile Behavior-based Networking Author Abdekrrahman Mtibaa
Faculty Advisor Khaled Harras, Ph.D.
Abstract Mobile devices are becoming ubiquitous, with ever increasing capabilities in terms of communication, computation, storage and sensing. The unprecedented tight coupling between mobile users and such devices provides promising opportunities to sense, model and accurately infer mobile user behavior and preferences. Future mobile services are expected to center around user preferences and social behavior. Hence, new ways of connecting mobile users are needed to support socially aware mobile networking. We explore leveraging social information to establish trustworthy and fair communication for mobile opportunistic networks. we investigate the potential impact of the lack of trust and fairness on node cooperation in such networks. We adopt a real-trace driven approach to study and analyze the trade-off between trust/ fairness and message success delivery rates in mobile opportunistic networks.
49
MoM_Digest_allPages.indd 54
4/20/11 3:04 PM
MoM_Digest_allPages.indd 55
4/20/11 3:04 PM
On Packet Loss and SINR in Wireless Local Area Networks Authors Eman Fituri Vinay Kolar
Faculty Advisor Nael Abu-Ghazaleh, Ph.D.
Abstract Rate adaptation techniques in wireless networks aim to opportunistically adjust the transmission rate of the users according to their channel conditions in order to maximize the overall network throughput. In general, existing rate adaptation techniques can be classified into two broad categories: loss based, where a transmitter adjusts the transmission rate based on the observed packet loss, and SNR based, where the transmission rate is chosen according to some SNR threshold. The SNR based techniques require training of the network in order to estimate these SNR thresholds, and it has been shown that these techniques can achieve higher throughput in static networks. However, mobility changes the channel model and the channel coherence time, which in turn renders the SNR thresholds obsolete. In this work, we design a rate adaptation protocol based on sequential hypothesis testing. This protocol estimates the packet loss rate and the SNR thresholds as the packets arrive sequentially. We implement this protocol on a WARP SDR board. Our initial experimental results show that this protocol can achieve high throughput under dynamic channel conditions.
51
MoM_Digest_allPages.indd 56
4/20/11 3:04 PM
MoM_Digest_allPages.indd 57
4/20/11 3:04 PM
Video Aggregation: High-Quality Video Streaming in Multi Camera Surveillance Systems Author Vinay Kolar
Faculty Advisor Nael Abu-Ghazaleh, Ph.D. Khaled Harras, Ph.D.
Abstract Video surveillance systems are becoming increasingly adopted in many applications such as border surveillance, home security, traffic surveillance and public-area monitoring. While these video streams provide invaluable feedback for monitoring activities, it presents an extremely challenging task in allocating human and network infrastructure. Human operators have to constantly monitor streams from hundreds of cameras, thus resulting in large personnel overhead. The network infrastructure should transfer the high-quality surveillance video streams from multiple cameras (typically in the order of couple of Mbps of data per video stream) round-the-clock; this causes network congestion, packet drops and, thus possible video quality degradation. However, most of the data in video streams is highly redundant; the streams often consist background scenes that do not change for long periods, and overlapping parts of the scene that is covered by multiple cameras. Hence, it is a waste of network and human resources to transmit and examine all the streams. In this study, we present a scheme called “Video aggregation� where the multiple video streams are intelligently aggregated by the intermediate network routers. Each router stitches multiple video streams that pass through it and forwards only the aggregated stitched video stream. Such aggregation scheme removes the overlapping redundant areas, and forwards only essential information towards the observer. The monitoring personnel is presented with a simple interface of one stitched view-of-the-world instead of multiple video streams, thus eliminating the need for monitoring and synchronizing multiple streams. Hence, the proposed scheme reduces the human costs in video surveillance and delivers maximum useful information supported by the network infrastructure.
53
MoM_Digest_allPages.indd 58
4/20/11 3:04 PM
MoM_Digest_allPages.indd 59
4/20/11 3:04 PM
C A R N E G I E
M E L L O N
U N I V E R S I T Y
I N
Q A T A R
P. O . B o x 2 4 8 6 6 | E d u c a t i o n C i t y, D o h a , Q a t a r | P h : + 9 7 4 4 4 5 4 8 4 0 0 | F a x : + 9 7 4 4 4 5 4 8 4 1 0 | w w w. q a t a r. c m u . e d u
MoM_Digest_allPages.indd 60
4/20/11 3:04 PM