Prediction Of Wrist Angle Under Different Loads Based On SEMG

In the field of surface electromyography(s EMG)pattern recognition,most of the current research focuses on qualitative recognition of static patterns of different limbs,aiming at specific tasks and actions of limbs.But in the process of practical application,in addition to knowing the movement of limbs,we also need to know the change of load intensity and the position information of joint angle movement.Traditional static qualitative pattern recognition can't adjust the movement amplitude and load intensity,so it can't accurately predict the joint angle.Therefore,it is of great significance to study the continuous prediction of wrist angle under different load intensity.In view of the above problems,this paper takes the wrist angle as the research object,uses the correlation between the surface EMG signal and the joint angle signal,and based on the EMG characteristics and neural network,carries out the wrist angle quantitative continuous recognition and wrist angle prediction under different loads.The main research work of this paper is as follows:Surface EMG signal preprocessing based on SNR coefficient evaluation model.Based on the analysis of the characteristics of s EMG signal,the noise that may be mixed in the wrist angle s EMG signal is determined.Using digital signal filtering algorithm,comb notch filter is designed to filter the power frequency noise.The evaluation model of signal-to-noise ratio coefficient is established,and the whole signal segment data is denoised by using wavelet filtering algorithm,which realizes the optimal processing of s EMG signal.Time-frequency feature combination based on feature correlation.By filtering the time-frequency features step by step,a feature set of EMG with maximum correlation is constructed.Principal component analysis(PCA)is introduced to reduce the dimension of feature combination signals,remove redundant channel features,and get the best EMG feature combination.The continuous prediction model of wrist angle under different loads based on optimization of extreme learning machine network by genetic algorithm.Aiming at the problem that qualitative pattern recognition can not adjust the wrist movement range and different load training intensity,the angle prediction model of the optimized extreme learning machine based on genetic algorithm is established to analyze the influence of different loads on the continuous prediction accuracy of wrist angle,so as to realize the accurate prediction of the continuous quantitative angle of wrist.Experimental analysis shows that the prediction effect of wrist angle becomes worse with the increase of load weight.Under the minimum load,the angle of wrist joint predicted by the limit learning machine optimized by genetic algorithm is closer to the actual angle,and the average error is about 5.94 °.