Research On The Processing Of Surface Electromyography Signal And Pattern Recognition

The surface electromyography signal (sEMG), which is acquired by using the surface electrodes. This is a kind of electric signal caused by the movement of several muscles and is obtained from human skins. Different movements will produce different electromyography signals and there is a lot of important information included in the electromyography signal. So surface electromyography signal plays an important role in the study of human action recognition, at the same time, the sEMG is widely used in the human body prosthetic control, rehabilitation training, clinical medicine, mechanical control and so on.This paper is to recognize the motion of human body by processing the sEMG signal, and then realize the control of prosthesis and help the disabled to do the rehabilitation training. This paper mainly studied the signal de-noising, feature extraction and pattern recognition. This is also the problem that many experts at home and abroad mainly concerned.The main work of this paper is:(1)We collect the sEMG to be distinguished and then analysis and process it. Then, according to its nonlinear and chaos characteristics, puts forward the optimal wavelet packet adaptive threshold de-noising algorithm, whether signal-to-noise ratio or mean square error, our method is better than the traditional wavelet de-noising algorithm.(2)The paper analyzes the wavelet coefficients, autoregressive model and other several classic methods for sEMG feature extraction. Combination with the characteristics of sEMG, we proposed two methods, which are the most absolute value of high frequency wavelet coefficient and the joint of wavelet coefficients and the largest lyapunov index. These two kinds of features have good recognition characteristics.(3)By using the back propagation(BP) neural network classifier realized the recognition of the human body gestures and the lower limb movement.Through the improvement of the criterion and optimize the implicit neurons number and the training error, the recognition rate is greatly improved and achieve better identification effect.This paper also designed the sEMG signal processing and pattern recognition software system, this system uses the proposed algorithm, reach the good recognition, we recognized the gestures of upper limb and lower limb, the average recognition is100%.