New approach of threshold estimation for denoising ECG signal using wavelet transform Harishchandra T. Patil Department of Instrumentation and Control Cummins College of Engineering for Women Karvenagar, Pune-411052 Email: harishchanddra.patil@gmail.com R. S. Holambe Department of Instrumentation and Control S.G.G.S. Institute of engineering and Technology Nanded-543243 Email: rsholambe@gmail.com Abstract—This paper presents a new method of threshold estimation for ECG signal denoising using wavelet decomposition. In this method, threshold is computed using the maximum and minimum wavelet coefficients at each level. Using this thresh- old and well known Hard thresholding process, the significant wavelet coefficients from each level are selected and denoised ECG signal is reconstructed with inverse wavelet transform. The performance of this method is compared with all well know wavelet shrinkage denoising methods with bior4.4 wavelet using root mean square error (RMSE) and signal to noise ratio (SNR) on MIT-BIH ECG database. The proposed threshold estimation is simple and faster compared to all existing threshold calculation methods namely VisuShrink, SureShrink, BayesShrink, and level- dependent threshold estimation and gives better SNR and RMSE. Proposed threshold estimation process decreases data sorting and storing resources allowing low-cost and faster implementation for portable biomedical devices. I. I NTRODUCTION This is the era of embedded devices. Most of the biomedical devices are portable. Power consumption of such battery operated devices is the critical issue [2]. That is why it is necessary to design biomedical devices with low power ratings. Remote ECG monitoring, know as Holter Monitor, is one of the portable devices, which sends ECG signal to health centre. In Holter monitor, ECG denoising is the critical issue. Large number of researchers have developed various ECG denoising methods [3] - [9]. Most of the ECG denoising methods are based on wavelet transform. Wavelet transform has been proved to be a successful tool for analysis of biosignals. Different wavelet based methods are used for denoising biosignals. Methods based on shrinkage of wavelet coefficients are very popular. In these methods [10]- [14] noisy signal is decomposed into wavelet coefficients by applying wavelet transform. After fixing the threshold using any of the method and using Hard or Soft threshold process, the significant wavelet coefficients are selected and the noise free ECG signal is reconstructed. Most of the ECG denoising methods which are based on wavelet transform, have used threshold estimation given by Donohos Universal theory, Steins unbiased risk estimator (SURE), Bayesian rule, etc [10]- [15]. All these threshold estimators required the mean, variance and median of signal, noise and wavelet coefficients. Calculating mean, variance and median of signal, noise and wavelet coefficients are more complex and time consuming in computing threshold. This is more time and power consuming in portable and wireless ECG monitoring systems. In real time ECG monitoring devices, the noise variance is unknown. Hence, the threshold estimation must be constantly updated. This paper presents a new approach of threshold estimation for denoising corrupted ECG signal using wavelet transform. Proposed threshold estimator is simple and level-dependent. The performance of the proposed method is analysed and compared with methods based on wavelet shrinkage [10]- [14] using MIT-BIH ECG database, contaminated with various types of noise. The paper is organized as follows: Section II presents a brief introduction to the ECG denoising, wavelet transform and performance metrics. In Section III, proposed threshold calculation and ECG denoising are elaborated in detail. Testing of proposed threshold estimator for ECG signal denoising on the MIT-BIH ECG database [1] and performance analysis are discussed in section IV. Finally, conclusion is given in Section V. II. MATERIAL AND METHODS A. ECG Denoising using wavelet transform Wavelet ECG denoising is done by following three basic steps. 1) Wavelet Decomposition: Noisy ECG signal is converted into wavelet domain. 2) Wavelet Denoising: Threshold is calculated by appropri- ate threshold rule and significant wavelet coefficients are selected by either hard or soft thresholding. 3) Wavelet Reconstruction: Wavelet reconstruction is done based on selected wavelet coefficients. B. Wavelet Transform Wavelet transform is widely used tool in biomedical signal processing. Wavelet functions [17], [18] are generated from mother wavelet () by scaling factor  and translation factor  :  , ()= 1 √  (  −   ) (1) 978-1-4799-2275-8/13/$31.00 ©2013 IEEE 2013 Annual IEEE India Conference (INDICON)