Research On Semg Signal Feature Enhancement And Classification Algorithm For Gesture Recognition

Surface electromyography signal(s EMG)is a potential signal generated by human muscle movement,which is closely related to muscle activity.To a certain extent,s EMG can reflect the functional state of neuromuscles,and easy to collect,good biomimetic,has been widely used in auxiliary diagnostic research.However,s EMG signals are highly susceptible to environmental influences,and there are problems such as weak robustness and unstable performance in actual scenarios.Therefore,in view of the problems of insufficient adaptability and low recognition accuracy of most existing methods,this paper carries out the following main work to realize the fast and accurate gesture recognition of s EMG signals:(1)In terms of s EMG signal preprocessing,aiming at the problem that the impact characteristics in the s EMG signal are easily drowned by the noise signal,a noise reduction method with zero treatment is proposed.The core of zeroing processing is to zero the data point whose absolute value is less than the zeroing threshold,and indirectly obtain the change of eigenquantity value by changing the signal sparsity.Experimental results show that the zeroing treatment can effectively remove noise interference,enhance the impact signal and improve the classification effect.(2)Aiming at the problems of excessive dimensionality and correlation of s EMG,this paper first selects and filters features through feature selection and fuses suitable features to construct an optimal feature set.Then,a feature enhancement method based on Gramian angular summation field-linear discriminant analysis is proposed,which can provide more feature information to characterize action features,and fuse the extended features with the original features to achieve the effect of in-depth mining of feature potential information.By comparing the results before and after feature enhancement,and by comparing the performance of the Gramian angular summation field-linear discriminant analysis algorithm with other feature processing methods,it is found that the data enhanced by the Gramian angular summation field-linear discriminant analysis algorithm perform better than the original data.(3)In order to better decode muscle activity information,aiming at the problem that the classical artificial bee colony algorithm has a slow convergence speed and is easy to fall into local optimum,the Gaussian variation operator with accelerated propagation speed and control convergence speed is introduced,and the classification model based on the improved artificial bee colony algorithm to optimize the support vector machine was constructed,and the research on gesture action classification recognition based on surface EMG signals was completed.Experimental results prove the stability and efficiency of the proposed method.This paper constructs a gesture recognition classification model based on s EMG signals,which provides theoretical support for high-precision motion intention analysis and deeper exploration of muscle neural control systems.