IJIRST –International Journal for Innovative Research in Science & Technology| Volume 3 | Issue 02 | July 2016 ISSN (online): 2349-6010
Wavelet based Intelligent Thresholding Techniques for Denoising ECG Signals Pushpanjali M P.G. Student Department of Digital Electronics & Communication Systems VTU-CPGSB VIAT, Muddenahalli, Bengaluru, India
Dr. Sarika Tale Associate Professor Department of Digital Electronics & Communication Systems VTU-CPGSB VIAT, Muddenahalli, Bengaluru, India
Abstract The electrocardiogram (ECG) signals are electrical signals produced by the heart during polarization and depolarization of the minute electric potentials that characterize the heart conditions and its rhythmic behaviors. These signals are processed and interpreted using electronic biomedical instrumentation techniques during which the ECG gets contaminated while acquiring the ECG signals leading to noisy signals due to the line power supply, leaky electrodes, signals processing stages and makes it difficult to read and interpret the ECG signals. From DSP point of view, the ECG signals are non-stationary which have sharp variations and high entropy. The DSP techniques based on the wavelets are powerful techniques that address the removal of noises due to their inherent separation properties based on the entropy and uncertainty properties of the ECG signals. Since entropy can measure the features of uncertainty associated with the ECG signal, a novel comprehensive entropy criterion and soft/hard thresholding techniques are proposed in which the coefficient of the wavelets are processed with built-in intelligent behavior and predict the best wavelet coefficients based on the multiple criteria related to entropy and energy. Taking account of the decomposition capability of wavelets and the similarity in information between the decomposed coefficients and the analyzed signal, the proposed novel intelligent thresholding using the both soft and hard thresholding is implemented using the criterion well as comparison information entropy for optimal wavelet selection leading to the intelligent thresholding and better performance. The experimental validation is conducted on the basis of ECG signals of 30 subjects selected from the MIT-BIH Arrhythmia Database. The performance indices is compared with each of these eight criteria through four filtering performance indexes, i.e., output signal to noise ratio (SNR), root mean square error (RMSE), percent root mean-square difference (PRD) and correlation coefficients. The filtering results of thirty ECG signals contaminated by noise have verified using the intelligent thresholding and performed better ECG signal filtering (10-20%) with hard and soft thresholding techniques. Keywords: ECG signals, denoising, filtering, DSP, wavelets, thresholding, performance modelling, pattern recognition, data mining and learning algorithms _______________________________________________________________________________________________________ I.
INTRODUCTION
Cardiovascular disease is one of the most causes of death in the world. With the aging trend of the population, people are paying more and more attention to research into telemedicine systems for the immediate and accurate detection of cardiac diseases. As a noninvasive test for recording the electric activity of the heart, electrocardiogram (ECG) plays a vital role in cardiac telemedicine systems [1]. The assessment of alterations in the features of ECG signals provides useful information for the detection, diagnosis and treatment of cardiac diseases. However, during the ECG signal acquisition and transmission procedures, the sampled ECG signal is inevitably corrupted by various noises, such as baseline wander, electrode motion, power line interference, motion artifact and other electronic disturbances. Usually, some specific measures such as median filter and band-stop filter can be implemented to suppress the influence of baseline wander and power line interference existing in ECG signals, respectively. However, electromagnetic disturbances such as thermal noise existing in measurement circuits have a significant influence on ECG signals. Thus, the noise reduction of ECG signals is a key requirement prior to pathological feature analysis II. PROBLEM STATEMENT The denoising of ECG signals is a persistent problem in the ECG contaminations and effecting the interpretations of ECG due to presence of noises corrupted by the various sources of signals at various stages of the ECG signal processing. The wavelet transforms are powerful techniques used in the noise removal and reductions of SNRs and other performance indices [2]. The thresholding techniques are common ways of removal of noises in which the local and global signal characteristics are explored. Hard and soft thresholding are two types of thresholding of ECG signals which are used for the noise removal applications. In particular, the soft thresholding of noisy ECG signals based on the best estimation of wavelet coefficient that helps in low frequency and high frequency ranges for the effective removal noises without losing the clinically important ECG signals.
All rights reserved by www.ijirst.org
379
Wavelet based Intelligent Thresholding Techniques for Denoising ECG Signals (IJIRST/ Volume 3 / Issue 02/ 061)
III. OBJECTIVES In this research work, we address the following DSP challenges for the noise removals from the contaminated ECGs signals. Design, implement and test the soft thresholding algorithms for various configuration and signal parameters such as noise level (dB), wavelet coefficient estimation etc. on MIT ECG databases. Develop a novel algorithm based on an intelligent soft thresholding in which the estimation of wavelet coefficients by applying the predication techniques are explored and estimated. Implement the proposed algorithm in MATLAB environment and derive the performance indices. Develop an intuitive GUIs and with effective user experience IV. WAVELET TRANSFORMS The wavelet transform have unique properties or characteristics which make useful and provide a powerful tool for their versatility and application domain [3]. The unique and important characteristics are: In both time and frequency the wavlets are localized. Extremely useful for the signals which are not stationary. Dynamic jumps, non-smooth features, applicable to the physiological signals (ECG, PCG, EMG, EKG etc.) As they fit in the characteristics of the wavlet transforms. Decomposition of a signal into multi-resolution components and characterized by “variable time frequency” resolutions is performed by wavlet transform V. DESIGN METHODOLOGY AND ITS IMPLEMENTATION The major design steps are explained as follows: 1) Noise Generation and Addition: An arbitrary noise is created [9] using Matlab library function used for the simulation of synthetic noise signals with known statistical properties. It provides tool for the noise generation that mimic the behavior of the noise in the ECG signals. In MIT ECG dataset, there are many ECG data files with noise contamination and collected while acquiring the ECG signals. The MIT ECG database provides a good collection of noise models and can be used for the experimentations of and performance analysis of denoising algorithms. It is represented by[10] S(n) + V (N) = Vs(n), wherever Vs(n) remains as the noisy ECG signal, V(N) remains free noise of the ECG, S(n) remains random high pass noise[9] 2) Signal Decomposition: The signals are disintegrated into noisy and original signals by means of the wavelet [9] transforms in to multilevel (e.g., level 5) by DWT using the Daubechies wavelets (Db4) from the wavelet libraries. We can use a family of wavelets depending up on the objectives of the experimentation. Thresholding: To each level of a threshold value is acquired from a hoop and applied meant for detailed coefficients of the noisy and original signals. The optimal threshold value remains estimated based on the minimum noise error criteria. In general, it is obtained by [9] captivating the minimum error amongst detailed coefficients of the noisy signals and those of original signals A hard or soft thresholding is utilized to shrinkage wavelets is applied so that d(c j, k) = 1 ………… Cj, k > T = 0 ………… otherwise Where d(c j, k) – output wavelet transform coefficients later thresholding, c j, k - wavelet transform coefficients, and T as the selected threshold based on the finest estimation [9]. 3) Signal Reconstruction: The original signal is reconstructed over inverse discrete wavelet transform (IDWT) and thresholding values at estimated values and may affect the morphology of the ECG signals. The best estimates of the coefficients comprise the low frequency of the original signal where the signal energy is concentrated. VI. PERFORMANCE ANALYSIS Image the signal S (n) = [s0, s1, s2, s3, ……. sN) is the original ECG signals which can be selected in the GUI and internally it maps to the MIT ECG database sets that denotes the original and denoised ECG files after applying the thresholds[10]. The filtering performances of different best wavelets are evaluated through comparing S and Sˆ based on four filtering indexes defined in the following The output signal to noise ratio SNRo
(1) The root mean square error RMSE is given by (2) The percent root mean-square difference PRD: All rights reserved by www.ijirst.org
380
Wavelet based Intelligent Thresholding Techniques for Denoising ECG Signals (IJIRST/ Volume 3 / Issue 02/ 061)
(3)
(4) It is noticed that the SNR is a measure of signal strength related to the background noise after wavelet thresholding denoising. Theoretically, the SNR of a filtered ECG signal should be large in amplitude in order to recover the useful signal. On the other hand, the RMSE reflects the distortion of the filtering result. The smaller the RMSE is, the closer the denoised signal is to the original signal and less distortion of the denoised signal after filtering [10]. Moreover, the PRD gives the information about the percentage of distortion of the filtered signal. A small PRD value indicates the efficiency of the denoising procedure. In addition, the correlation coefficient r is a statistical concept that measures how well the denoised signals follow the actual signal. The larger value of r is, the closer the denoised signal is to the original signal VII. EXPERIMENTAL RESULTS The various reference ECG data sets were tested with the thresholding by hard and soft levels using the algorithms developed. The noise data with various values are depicted in the following diagram
Fig. 1: Noisy and Denoised ECG signal for cu01.dat
Fig. 2: Noisy and Denoised ECG signal for cu02.dat
Graph of threshold for various values of the ECG dataset from the reference datasets are represented below and values are tabulated
All rights reserved by www.ijirst.org
381
Wavelet based Intelligent Thresholding Techniques for Denoising ECG Signals (IJIRST/ Volume 3 / Issue 02/ 061)
VIII. CONCLUSION The wavelets transforms have been efficient denoising analysis tools for the time varying and non-stationary signals such as ECG signals. The choice of best wavelet is critical issues in wavelet filleting and thresholding of ECG signals. The ECG signals of MIT ECG dataset with specific reference to the MIT-BIH Arrhythmia Databases with noise contamination in various noise interferences scenarios using thresholding found to be useful with better performance indices The thresholding values, estimation of wavelets coefficients and making a calculated estimation with a small given set of data and predicting for small range of the signals found to be effective with 80% accuracy. However, it is difficult to estimate the global parameters with local information. The proposed method makes a short term predication and helps in noise reductions by 65-78% and needs further investigations. Therefore, our research in the future will concentrate on the progress of complete entropy criterion for different accidental pathological ECG signals. REFERENCES Dipti Thakur, Sagar Singh Rathore, “Comparison of ECG Signal Denoising Algorithms in Fir and Wavelet Domains” International Journal of Engineering Research and General Science Volume 3, Issue 6, November-December, 2015 ISSN 2091-2730 [2] Carmona R.A., Torresani B., “Characterization of Signals by the Ridges of Their Wavelet Transforms”, IEEE Transactions on Signals Processing, Vol.45, No.10, Oct.1997. [3] Su, L., Zhao, G., “De-noising of ECG signal using translation-Invariant wavelet de-noising method with improved thresholding,” Annual International Conference of the IEEE Engineering in Medicine and Biology - Proceedings 7 Volume, 2005, pp. 5946-5949. [4] Donoho, “De-noising by soft thresholding”, IEEE Trans. Trans. Inform. Theory, vol. 41, pp. 613-627, 1995. [5] Harishchandra T. Patil, R.S. Holambe, “New approach of threshold estimation for denoising ECG signal using wavelet transform”, India Conference (INDICON), 2013 Annual IEEE Conference, Mumbai, India, 13-15 Dec. 2013. [6] Upasana Mishra, Mr. Love Verma “Denoising of ECG signal using thresholding techniques with comparison of different types of wavelet”, ISSN 22771956/V2N2-1143-1148 [7] Chandrakar Kamath “Entropy-Based Algorithm to Detect Life Threatening Cardiac Arrhythmias Using Raw Electrocardiogram Signals” Middle-East Journal of Scientific Research 12 (10): 1403-1412, 2012 ISSN 1990-9233 © IDOSI Publications, 2012 [8] MM Elena, JM Quero, I Borrego University of Seville, Seville, Spain “An Optimal Technique for ECG Noise Reduction in Real Time Applications” ISSN 0276–6547 [9] Alfaouri, Mikhled. "ECG Signal Denoising By WaveletTransformThresholding”, American Journal of Applied Sciences/15469239, 20080301 Publication [10] He, Hong, Yonghong Tan, and Yuexia Wang."Optimal Base Wavelet Selection for ECG Noise Reduction Using a Comprehensive Entropy Criterion", Entropy, 2015. [1]
All rights reserved by www.ijirst.org
382