Detection And Pattern Recognition Of SEMG Based On Prosthesis

Surface EMG(Surface electromyography signal, s EMG) is an electrophysiological signal formed during muscle activity. In the field of surface EMG, decomposition and identification of human action based on surface EMG is a major focus of research. Through research, scientists are try hard to create a more convenient, natural, simple, effective and rapid human-computer interaction for human beings, which has far-reaching significance for sign language recognition, games and entertainment products, prosthetic control, operational command, mobile device control and sports electronics, etc.In the field of multi-function prosthetic control, surface EMG pattern recognition is a fundamental problem. This paper takes pattern classification means and extracting of EEG signal characteristics for further exploration and analysis. EMG signal mainly comes from four different surfaces electrodes of the forearm muscle tissue, select improved support vector machine as the pattern recognition method to identify different gestures.This pattern covering pre-processing, signal acquisition and hand movements identifying.This paper first take a comparative analysis for different feature extraction patterns, including frequency domain, time domain, wavelet features and many other EMG feature recognition means, take the wavelet packet energy value for dimension reduction, obtained the optimal feature vector.In pattern recognition, fist select support vector machine(SVM) and least squares support vector machine(LSSVM) from numbers of classifiers, and then input the acquired feature vectors to the classifier, after analysis and processing, take different surface electromyography signal of different hand gestures as the identify object, finally carrying on the classification process. In the process of classifier optimization, introduced genetic algorithm and particle swarm optimization for comparative tests, after the analyzing of training time and motion recognition rate we can found that for the four hand movements identification in this paper, based on the particle swarm optimization parameters of LSSVM pattern has a shorter operation time and higher recognition rate.