The Analysis Of Surface Electromyographic Signal Based On Wavelet Transform And Artificial Neural Network

EMG is the independent movement of human neuromuscular issued in biological signal, which reflects the nerve, muscle function.EMG is easy and painless electrodes measured the surface EMG signals. On awareness and understanding of human nervous system information transfer, basic medical research and clinical diagnosis, sports medicine and rehabilitation engineering are widely used. How do the surface EMG signal to extract information effectively and to achieve accurate, real-time action recognition, the surface EMG of the key issues. In this paper, the Canadian production of bio-feedback machine INFINITI multi-channel surface EMG signals collected for the study. The bio-feedback machine as a new type of EMG signal acquisition device, there are good prospects. But its acquisition of surface EMG signal analysis work related to relatively rare.Based on bio-feedback machine INFINITI Telecom upper surface of muscle action recognition research, the following major work done:1. Noise in the surface EMG signals were also studied.Soft threshold denoising method, discuss the threshold selection problem.Stratification of the Birge-Massart threshold and global threshold wavelet denoising methods were compared, stratified by the measured data show the threshold method.2. EMG for the non-stationary characteristics of wavelet transform method for feature extraction surface EMG signal, extract the maximum absolute value of wavelet coefficient feature vector, enter the neural network classifier for pattern recognition3. This innovative introduction of a Resilient back-PROPagation algorithm improved BP network, the improved BP network to identify the action surface EMG signal, the received classification speed, high precision of the BP category recognition network recognition accuracy as high as92.5%.