Surface electromyogram signal estimation based on wavelet thresholding technique.

Surface Electromyogram signal collected from the surface of skin is a biopotential signal that may be influenced by different types of noise. This is a considerable drawback in the processing of the sEMG signals. To acquire the clean sEMG that contains useful information, we need to detect and eliminate these unwanted parts of signal. In this work, a new method based on wavelet thresholding technique is presented which provides an acceptable purified sEMG signal. sEMG signals for this study are extracted for various hand movements. We use three hand movements to calculate the near optimal estimation parameters. In this work two types of thresholding techniques, namely Stein unbiased risk (SURE) estimator and adaptive Bayes estimator are utilized coupled with selected types of mother wavelets with different levels of decomposition. After designing the estimation technique, for evaluating the efficacy of method, the formed signals are sent to a pattern recognition system in order to discriminate among eight hand movements. The acquired results indicate that the wavelet based estimation technique using SURE thresholding approach is an appropriate method for producing sEMG signals without noise that may result in considerable improvement in the application of hand movement recognition.