A Novel Technique of ECG Denoising based on Lifting Wavelet Transform and Total Variation Minimization

Abstract

Buckground:The signal of Electrocardiogram (•) is one of the most popular diagnostic means providing an electrical picture of the heart and also information about different pathological conditions. Due to the path deformities and external electrical disturbances, the signal of becomes noisy. Hence, in literature, many ECG denoising algorithms have been proposed and among them we can mention the techniques based on wavelet coefficient shrinkage. The purpose of this paper is to denoise ECG signals applying a new ECG Denoising technique and proving its performance compared to some denoising approaches existing in literature. This new proposed technique of B Denoising consists at the first step in applying the Lifting Wavelet Transform (u) to the noisy c Signal (where 2 is the decomposition level) in order to obtain three noisy wavelet sub-bands, k, g and r. The two coefficients, o, u are details ones and they are denoised by soft or hard thresholding in order to obtain denoised coefficients, n1 and d2. The coefficient : is an approximation one and is denoised by Total Variation Minimization in order to obtain a denoised one, T2. Finally, the inverse of his applied to e, and s in order to obtain the denoised i signal. The evaluation of this proposed technique is performed by comparing it to three other denoising approaches existing in literature. The first one of these approaches is based on g, the second one is n double-density complex a denoising method and the third one is based on non local means.Results:All These techniques are applied on a number of lsignals taken from database and corrupted by an additive White Gaussian noise at different values of Signal to Noise Ratio (o). The obtained results from the computation of the f and the Mean Square Error ( ), show that the proposed technique outperforms the other three mentioned techniques.Conclusion:In this paper, the proposed ECG denoising technique based on E and l, outperforms the other previously mentioned denoising approaches and this based on the computation of the SNR and MSE.