Research On Recognition Method Of Hand Emg Signal Based On Exoskeleton Power

With the development of economy,people's pursuit of work efficiency has become stronger.When people are engaged in heavy physical work for a long time,it will cause arthritis,lumbar muscle strain,femoral head necrosis,frozen shoulder and other problems,which will seriously affect human health and work efficiency.The use of the EMG signal acquisition system to control the auxiliary work of the assisting device when the staff is carrying heavy objects not only solves the problem of work efficiency,but also reduces physical exertion.For this reason,this article has launched a study on the recognition of hand EMG signal assisted by exoskeleton.First,the acquisition method of EMG signal is researched.Through filtering to eliminate the noise generated by power frequency signals,skin sweat and other factors,so as to obtain effective electromyographic signals;through the microcontroller STM32 single-chip microcomputer to achieve the acquisition and transmission of myoelectric signals,complete the acquisition of myoelectric signals.Secondly,collect the EMG signals of five different gestures and preprocess the collected EMG signals.Analyze the data distribution of the EMG signal in the acquisition process,and detect the active segment through an improved moving window;select root mean square(RMS),zero crossings(ZC)and variance(VAR)to extract the features of the active segment,and cooperate Standard min-max normalization and index order scramble processing to realize the preprocessing of EMG signal.Then,the BP neural network algorithm is used to deal with the problem of non-linear correlation of the EMG signal caused by the correlation between different actions.According to the action of the EMG signal and the number of eigenvalues,the input layer,hidden layer and output layer of the classification model are determined.Select the Sigmoid excitation function,cross-entropy loss function,and Adam optimizer to optimize the training process of the classification model.Import the preprocessed feature data set into the classification model to complete the training of the classification model.Finally,based on the PyQt5 environment,complete the interface development of the upper computer test system,and design the corresponding experimental platform;establish a control system to complete the real-time communication between the upper computer and the lower computer;test five different gesture actions to control the stepping motor to complete the corresponding steering and Rotation speed verifies the reliability of the EMG signal test system.