Nonlinear Characteristic Research Of Action Surface Electromyography Signal

Human body action electric signals are generated by motion units which are composed of nerves and muscles and then control muscles to complete muscular synergies and realize human body action. Different electric signal drives different action. During action realization, these electic signals are output on skin surface through human tissues and are acquired by services on the skin such as surface electrode. The acquired signals are named as SEMG (surface electromyography, SEMG) signals. SEMG signals have different association with muscular activity and fuction, and reflect the situation of nerves and muscles and their activities. Therefore, SEMG signals are widely used in Clinical Medicine, Sports Medicine, Ergonomics, Rehabitation Medicine, Neurophysiology and Electrophysiology.Currently, nonlinearity research about action SEMG signals is in the initial stages of exploration. At the existing basis of SEMG signals acquisition motheds and technologies, experiment acquired SEMG signals was designed, and six kinds of action SEMG signals were acquired as research object from human forearm. The six kinds of action include wrist flexion, wrist extension, hand grasp, hand extension, wrist flexion up and wrist flexion down. Nonlinearity of SEMG signals was studied and verified using nonlinear time series analysis methods, demonstrating that action SEMG signals are chaotic. The results have very important values on deeply recognizing law and essence of nerve and muscle function, establishing more scientific and rational muscle function non-invasive evaluate techniques.For learning more about nonliearity of action SEMG signals, Wavelet Transform and Hilbert-Huang Transform were used in multi-scale decomposition of surface SEMG signals, and then studied nonliearity of SEMG signal on every scale, increasing understanding about nonlinearity of SEMG signals by extanding nonlinearity on different scales.In the last, for the purpose of increasing the recognition accuracy of action SEMG signals, a method combining nonlinear analysis with multi-scale analysis is proposed. Considering the nonlinear and non-stationary characteristic of SEMG, multi-scale nonlinear features were introduced and applied to the pattern recognition of six types forearm action SEMG signals. Combining kernel principle component analysis, multi-scale nonlinear features were input into support vector machine for recognition. The mean recognition accuracy reaches 98%. The results show that the method applying multi-scale nonlinear features in the pattern recognition of action SEMG signals is effective and precise.