Comparative study of denoising sEMG signals

Abstract

Denoising of surface electromyography (sEMG) signals plays a vital role in sEMG-based mechatronics applications and diagnosis of muscular diseases. In this study, 3 difierent denoising methods of sEMG signals, empirical mode decomposition, discrete wavelet transform (DWT), and median filter, are examined. These methods are applied to 5 diferent levels of noise-added synthetic sEMG signals. For the DWT-based denoising technique, 40 diferent wavelet functions, 4 diferent threshold-selection-rules, and 2 threshold-methods are tested iteratively. Three diferent windowsized median filters are applied as well. The SNR values of denoised synthetic signals are calculated, and the results are sed to select DWT and median filter method parameters. Finally, 3 methods with the optimum parameters are applied to the real sEMG signal acquired from the exor carpi radialis muscle and the visual results are presented.