A Weighted Bio-signal Denoising Approach Using Empirical Mode Decomposition

Purpose The purpose of this study is to show the effectiveness of a physiological signal denoising approach called EMD-DWT-CLS. Methods This paper presents a new approach for signal denoising based on empirical mode decomposition (EMD), discrete wavelet transform (DWT) thresholding, and constrained least squares (CLS). In particular, the noisy signal is decomposed by empirical mode decomposition (EMD) to obtain intrinsic mode functions (IMFs) plus a residue. Then, each IMF is denoised by using the discrete wavelet transform (DWT) thresholding technique. Finally, the denoised signal is recovered by performing a weighted summation of the denoised IMFs except the residue. The weights are determined by estimating a constrained least squares coefficients; where, the sum of the coefficients is constrained to unity. We used human ECG and EEG signals, and also two EEG signals from left and right cortex of two healthy adult rats. Results The 36 experimental results show that the proposed EMD-DWT-CLS provides higher signal-to-noise ratio (SNR) and lower mean of squared errors (MSE) than the classical EMD-DWT model. Conclusions Based on comparison with classical EMD-DWT model used in the literature, the proposed approach was found to be effective in human and animal physiological signals denoising.