Research On Recognition Method Of Motor Imagery Eeg Signal And Control Of Bionic Manipulator

Brain computer interface(BCI)can control medical devices,wheelchairs,manipulators,smart home and other external devices to express brain activity information,and achieve the purpose of restoring or even enhancing human physical ability.However,the research and development of BCI technology and system is still in the laboratory stage,and few BCI systems have been put into practical application.This is because the accuracy and complexity of recognition algorithm restrict the realization of real-time BCI system.Based on this,this paper studies the motor imaging(MI)EEG signal recognition method and bionic manipulator control,aiming to realize the interaction between brain and bionic manipulator.The main research results are as follows:(1)Research on experimental paradigm.Aiming at the problem that multi degree of freedom motion of manipulator needs more control signals,based on the two classification and three classification motor imagination,a two-level motor imagination experimental paradigm is designed,which includes the two classification of left and right hand and the three classification of hand grasping,wrist flexion and elbow flexion.EEG signals of 12 subjects were collected as data sets.(2)Preprocessing and feature extraction methods.In order to solve the problem of weak amplitude and interference of EEG,Butterworth band-pass filter,common average reference(CAR)and independent component analysis(ICA)are used to reduce the noise of EEG in time,frequency and spatial domain to improve the signal-to-noise ratio.For common spatial pattern(CSP),the maximum entropy principle is introduced on the basis of regularized reconstruction of covariance matrix to reduce the ill conditioned effect of small samples on covariance matrix.The classification results of the same pattern recognition algorithm prove that empirical mode decomposition(EMD)combined with maximum entropy common space pattern has better feature extraction ability.(3)Research on pattern recognition method.In order to classify the left and right hand movements,a binary tree twin support vector machine model was constructed to classify the EEG signals.Whale optimization algorithm(WOA)is used to optimize the penalty factor of twin support vector machine(TWSVM),and compared with grid search(GS)twsvm in recognition rate.Experimental data show that woa-twsvm has better classification results than GS twsvm in pattern recognition.This paper further compares woa-twsvm algorithm with other classification methods,and the experimental results show that woa-twsvm achieves the highest recognition rate.(4)Research on bionic manipulator control.The control system of bionic manipulator is studied in software and hardware respectively.The hardware structure of bionic manipulator system based on STM32F103C8T6 processor is designed.The functions of grasping,wrist bending and elbow bending are realized.In the software,the task of motor imagery signal processing and control signal generation is completed based on Matlab GUI programming.In order to test the performance of the bionic manipulator control system,the second motor imagery task experiment is designed and collected.The experimental results prove the recognition accuracy of the proposed EEG signal processing algorithm and the effectiveness of the bionic manipulator control system.