Research On The Processing And Recognition Of Human Hand Actions SEMG Signals

Surface electromyography(EMG) signal is a complex weak bioelectric signal, which is collected from the surface of skin through the surface electrode and related instruments and can reflect the functional status of human nerves and muscles. With the development of society and the improvement of people's living standard, there is an increasing demand on the medical system with the function of intelligent rehabilitation. So it is one of the indispensable procedures to create the system by recognizing the sports consciousness of people accurately by the s EMG. In order to achieve the accurate recognition of different movements, the s EMG signal of human hand actions is studied in this paper. The major work as follows:First, the generation mechanism and the common characteristics of s EMG signal are introduced. The choice of hand actions and the EMG electrodes placement are confirmed. The s EMG signal acquisition system is built by choosing the appropriate hardware and software. And the two-channel s EMG signal acquisition of four different hand movements is completed.Second, in order to extracting the s EMG signal effectively in noisy background, the s EMG signal is de-noised by combined the empirical mode decomposition(EMD) and the wavelet threshold. Firstly, the white noise added the s EMG signal is decomposed into a sum that is made of a series intrinsic mode functions(IMF) with different frequency components. Then, the fitted IMF of high frequency is processed using wavelet threshold method. Finally, the de-noised IMF and all untreated IMF are fitted. Then these fitted signals are the surface electromyography signal which de-noised by the wavelet threshold and EMD. The experimental results show that the method combining the advantages of wavelet threshold and EMD has the better de-noising effect than others.Thirds, the characteristics of time domain, frequency domain and time-frequency of the de-noised s EMG signal are calculated respectively and compared analyzing. The feature of s EMG signal using wavelet transform method based on time-frequency analysis is ascertained. And the feature vector is consisted of the maximum absolute value of wavelet coefficients from four actions surface EMG signal.Finally, the EMG signal from four kinds of hand actions is classified by using support vector machine. And the artificial fish swarm algorithm is used to optimize the parameters of the support vector machine. The recognition accuracy of the model is improved by choosing the optimal parameter combination. Experimental results show that this method has the higher recognition rate.