Research On Gesture Recognition Algorithm Using Surface Emg Signal Based On Deep Learning Model

Surface electromyography(sEMG)is a muscle motor bioelectrical signal recorded by a surface electrode.The surface EMG signal can reflect the functional status of the nerve and muscle contraction.With the development of science and technology,researchers at home and abroad have gradually studied the recognition of surface EMG signal gestures.At present,they have made certain progress in sports training,rehabilitation training,clinical application and sports training.Along with the sudden emergence of Deep Learning,the Deep Learning model has significantly improved the accuracy of surface EMG signal gestures.Deep Learning,especially Deep Neural Networks,has a strong application value for building multi-modal perceptual computing models,but it is also requires to have enough data volume as support.At present,the precision prosthesis on the market is generally higher,and the laboratory electromyography treatment method is mainly tested on the computer simulation platform,and there is still a certain gap from the actual use.In order to solve the above problems,this dissertation proposed a surface EMG signal gesture recognition algorithm based on Deep Learning model,which can realize the recognition of surface EMG signals,which can be used for high precision intelligent prosthesis in the future.First of all,this dissertation used the Swiss Ninapro public database to preprocess the surface EMG signal.Since the original acquired surface EMG signal has a lot of noise interference,the waveform characteristics of surface EMG receive a certain degree of interference,which affects the accuracy of the final recognition effect.Therefore,this dissertation used a 2nd order Butterworth band rejection filter to remove noise interference from the surface EMG signal.Then,during the experiment,it is found that the original surface EMG signal has a very low amplitude of some bands,which makes the experimental results have a large deviation.Therefore,the standard deviation of the pre-processed surface EMG signal is used to filter out the no-signal segment.Ensure that each segment of the signal has the original characteristics of the surface EMG signal for optimum recognition accuracy.At the same time,the existing surface EMG signal are extended by overlapping time windows to meet the requirements of the Deep Learning model for data samples.Finally,the surface EMG signal is identified by a Deep Learning model.Since the surface EMG signal is a one-dimensional signal,the pre-processed data is trained and tested by a 32-layer ResNet convolution neural network by constructing a one-dimensional convolution neural network,and the experimental results are obtained.The algorithm of this dissertation is compared with the existing face recognition algorithm based on surface EMG signal.The results show that the surface electromyography signal gesture algorithm based on Deep Learning model effectively improves the recognition accuracy and develops accuracy.Highly robust intelligent prosthetics provide technical support.