Performance Comparison and Applications of Sparsity Based Techniques for Denoising of ECG Signal

Abstract

ECG denoising using different kinds of scientific techniques and methods has been an interesting research area among the signal processing research fraternity. There are various kinds of noises that interfere with ECG signal at different levels. Powerline interference, baseline wander noise and electromyography noise are at highest priority to remove from the desired signal. Several sparsity based adaptive and wavelet digital filtering techniques have been proposed in previous investigations for denoising of ECG signal. But there qualitative and quantitative performance analysis against each other is lacking in the literature. In this paper, we reviewed various sparsity based noise reduction techniques of adaptive and wavelet algorithms. Using the benchmark dataset of MIT/BIH, a detailed and fair comparison of LMS, RLS and DWT were implemented for their performance analysis. The qualitative analysis has been presented in terms of the morphology differences in the denoised signal and the quantitative analysis is presented in terms of various performance measuring parameters of signal-to-noise ratio (SNR), mean square error (MSE), percentage root mean square difference (PRD) and peak-signal-to-noise ratio (PSNR). The obtained results show that adaptive filtering using RLS algorithm performs better in more dense noisy conditions whereas the wavelet filtering is better to perform in less noisy conditions. Further, all three algorithms were tested on different kinds of noises like power-line interference, baseline wander and abrupt shift in the ECG data, where, DWT based filtering approach was found superior on removal of powerline and baseline wander interferences, but it fails to remove the abrupt shift kind of noise. The abrupt shift noise was best removed by both LMS and RLS adaptive algorithms but at the cost of low speed and poor quality. Thus, the presented optimized analysis of advanced three sparsity based filtering techniques would provide great potential benefits in biomedical applications of ECG signal processing, feature extraction, analysis and other related fields.