EEG-EMG Signal Fusion Analysis Based On Upper Limb Movement

The electroencephalogram(EEG)of the cerebral cortex reflects the body’s motor control information.The electromyographic signal(EMG)of the body’s muscle tissue reflects the functional response of the muscle to the brain’s control intention.The intersection of two physiological electrical signals has become a new hot field in artificial intelligence,medical rehabilitation and neuroscience.At present,most of the prostheses use EMG signals as control information,but some EMG signals are very weak and complex,and the motion recognition rate is low,which can not meet the actual application needs.Considering the patients with dyskinesia and normal EEG signals,based on the Brain-Computer Interface(BCI)technology,the feature fusion strategy of EEG signals and EMG signals can be used as important control information for rehabilitation prostheses.In order to improve the recognition rate and control accuracy of prosthetics in brain-computer interaction control,this thesis studies the fusion of EEG and EMG signals.First of all,this thesis describes the research background of EEG and EMG signals in the field of prosthetic control,and reviews the application status and existing problems of brain EMG signal fusion analysis.Based on the accuracy of neuromuscular function fusion and the control of EEG signals,the EEG and EMG signals were analyzed to improve the stability of the auxiliary equipment.According to the physiological basis and characteristics of the motor EEG signal and the EMG signal,the acquisition position of the experimental signal is determined.Secondly,the upper limb motor EEG and EMG signal acquisition experiments were designed to preprocess the acquired signals.Based on the wavelet threshold denoising method,an improved threshold algorithm is proposed to remove the noise in the EEG signal and improve the signal-to-noise ratio of the EEG signal.On the basis of the wavelet analysis method,the collected EMG signal is denoised by using the layered threshold denoising method.Then,the eigenmode function component of the EEG signal is obtained by empirical mode decomposition,and the feature of the motion EEG signal is extracted by Hilbert-Huang Transform(HHT)method.According to the characteristics of EMG signal,based on time-frequency domain analysis,wavelet transform is applied to extract the EMG characteristic values with high correlation.Finally,the sparse autoencoder is used to fuse the characteristics of EEG and EMG signals,and the support vector machine is used to classify and recognize the fusion signal and single motor EEG signals.The results show that the proposed method improves the recognition accuracy of motion.