An Experimental Investigation of Wavelets for ECG Signal Denoising

Abstract

The main aim of the presented paper is to explore all types of all wavelet families at all possible decomposition levels to find the best decomposition level as well as the best wavelet family to denoise an ECG signal from a specific type of noise which is the white Gaussian noise. The universal hard and soft thresholding of Donoho & Johnston is used to shrink extracted details. The MIT-BIH signal database is used to test our programs. The experiments revealed that the Most interesting wavelet level decomposition is level 5, and, Daubechie Family has proved its performance in denoising ECG signal when comparing the Signal to Noise Ratio (SNR) measure. These two findings are well confirmed by an expert in the domain.