Wavelet Based Feature Extraction of Electromyogram Signal for Denoising

Electrical signals recorded from muscles require processing before its use, so the modeling of these bioelectric signals is necessary. Wavelets are used for the processing of signals that are non-stationary and time varying. The surface Electromyogram signals were estimated with following steps, first, the obtained signal was decomposed using wavelet transform; then, decomposed coefficients were analyzed by threshold methods. With the appropriate choice of wavelet, it is possible to remove interference noise effectively in order to analyze the signal. This paper presents a comparative study of different Daubechies wavelets (db2-db14) family for analysis of arm motions. From the analyzed results, it was inferred that wavelet db4 performs denoising best among the wavelets and is suitable for accurate classification of surface Electromyogram signal. Because of the wavelet denoising, accurate observation of activity that is not possible with conventional filtering, becomes possible.