Comparative analysis of filters for cancellation of power-line-interference of ECG signal

Abstract

Filtering noises/artifacts from the electrocardiogram (ECG) can sustain the efficient clinical decision making. Comparative analysis of several filtering techniques is proposed: two adaptive noise cancellation techniques, Least Mean Square (LMS), Recursive Least Square (RLS); Savitzky-Golay (SG) smoothing filter and Discrete Wavelet Transform (DWT). These methods are implemented on 60 Hz Power-Line Interference (PLI), ECG signals of FANTASIA database and MIT-BIH Arrhythmia Database. Here, Short-Term Fourier Transforms (STFT) and Continuous Wavelet Transform (CWT) is introduced as a graphical tool to measure the noise level in the filtered ECG signals and also to validate the filtering performances of the proposed techniques. Statistical evaluation is also performed calculating the Signal to Noise Ratio (SNR), Mean Square Error (MSE), the Root Mean Square Error (RMSE), Peak Signal to Noise Ratio (PSNR) and Peak to Peak Amplitude (P2P) change before and after filtering of the ECG signals. The graphical results (frequency domain analysis using STFT and CWT) and statistical observation suggest that the noise cancellation performance of DWT is better, over other techniques.