Comparative Study of ECG Signal Denoising by Empirical Mode Decomposition and Thresholding Functions

Electrocardiogram (ECG) signal is widely used for diagnosing cardiac diseases. Several denoising methods have been proposed based on Empirical Mode Decomposition (EMD). Moreover, EMD is a successful tool for denoising. In this paper a review of comparative study of ECG signal denoising based on EMD and Thresholding Functions is presented. Five denoising algorithms (EMD-Conv, EMD-IT-Soft, EMD-IT-Hard, EMD-ITF and EMD-Custom) are applied on real ECG signals contaminated with different levels of white gaussian noise. EMD was applied to decompose adaptively a noisy signal into Intrinsic Mode Functions (IMFs). The noisy IMFs were denoised by thresholding functions. The performances are evaluated by measuring signal to noise ratio in dB and mean square error (MSE). EMD-Conv, EMD-IT-Soft, EMD-IT-Hard are used as references techniques. Simulation results show that the EMD-ITF and EMD-Custom approaches outperforms the conventional EMD methods.