Rehabilitation Manipulator Control Based On Fusion Of Eeg And Emg Signal

The number of stroke patients in my country is increasing year by year,and most of them have sequelae of hand motor dysfunction.Based on the motor imagery braincomputer interface system and the electromyography interface system,it can control the rehabilitation manipulator,help stroke patients to carry out active rehabilitation training,promote the self-healing of damaged nerves,and restore the functional movement ability of the hand.Due to the weakness and complexity of EEG signals,the motor imagery brain-computer interface system takes a long time to train,has poor data reusability,and has low recognition accuracy.The rehabilitation manipulator controlled by the EMG interface also has accurate recognition in the multi-task classification of hand movements.low rate problem.This paper studies a hybrid brain-computer interface system based on EMG fusion,which utilizes the synergistic complementarity between EMG signals to improve the accuracy of motion pattern classification.On the one hand,transfer learning is introduced into the processing of motor imagery EEG signals,which reduces the training time of the EEG signal classification model and improves the recognition rate of EEG signals.On the other hand,a hybrid neural network model is proposed to improve the recognition accuracy of EMG signals under multi-classification tasks of hand movements.Finally,the classification results of EEG signals and EMG signals are fused to obtain recognition results with higher classification accuracy.Rehabilitation manipulator for motion control.To meet the needs of hand motor function rehabilitation training for stroke patients,the main tasks of this paper are as follows:Firstly transfer learning is introduced into the processing of motor imagery EEG signals,and in response to the problems of poor migration effects that exist solely based on instance migration and feature-based migration,a hybrid migration learning method combining instance migration and feature migration is proposed,with the first step based on the Euclidean spatial data alignment method for instance-level migration to reduce the distribution differences between source and target domain data The second step improves the CMMS method based on the idea of minimizing the maximum mean difference,completes the screening and reconstruction of the source domain features,completes the migration based on the special level,and improves the migration effect.And the BCI competition dataset is used for the validation of the method.Secondly,aiming at the problem of low recognition accuracy of traditional machine learning and neural network in the four-classification task of hand motion of EMG interface,a neural network combining multi-scale convolutional neural network(MSCNN)and long short-term memory network(LSTM)is proposed.The network model first uses MSCNN to fit the characteristics of EMG signals from different scales,and then enhances its time series features through LSTM network to improve the recognition accuracy of the four classifications of hand motion at the EMG interface.Finally,the Bayesian algorithm used in EMG decision fusion is introduced,and the control system of rehabilitation manipulator for EMG fusion is designed and developed,including signal acquisition system,upper computer stimulation system,manipulator and control system,and the research method in this paper is embedded in the control system.In the system,the experimental paradigm of brain EMG signal acquisition of hand grasping and stretching actions is designed and verified by online experiments.