Research On Feature Extraction Method For ASEMG

Action Surface Electromyographic is a kind of complex bio-medical signal. It is the electric fluctuation of the shrinkage of muscle. It is not only related to the tissue and physiological characteristics of muscle, but also connected with the nervous controlling system. Therefore, ASEMG has been the hotspot of many areas, such as clinical diagnosis, rehabilitation medicine, bio-mechanical.Focus on the feature extraction, my research focus on how to extract effective features to quickly and accurately recognize different action patterns. There used many signal processing methods in the paper, such as time domain, frequency domain, time-frequency domain and the view of nonlinear to do feature extraction. Then I used BP neural network to do the pattern recognition. Also lots of comparison in transverse and vertical of different signal processing methods frequently used is done. Also much effort was given to trying different methods. Finally, based on the existed methods, some new and unique methods, such as Nonlinear characterizes combined with EMD, energy distribution with Euclidean space and joint feature identification were put forward, and gained good results, which proposed some new research methods and different thinking for improving the research of motion ASEMG, so does to the application.