Research On Motor Rehabilitation Training System Based On Motor Imagery Eeg Signals

Motor imagery(MI)is a dynamic process of rehearsal action on the memory of the scheduled presentation,without actual limb movement.Brain-Computer Interface(BCI)based on motor imagery EEG is a channel established between brain and computer or other devices using motor imagery,without the need to use language and body movements.This technique can be well used in patients with paralysis and stroke for rehabilitation.A number of studies have shown that with the increase of motor imagery tasks,the classification accuracy will be reduced.Therefore,it is necessary to study the multi-class motor imagery.Based on the research of feature extraction algorithm using common spatial pattern and feature selection algorithm applying firefly algorithm,improved algorithms are proposed.Finally,a motor rehabilitation training system is designed and implemented.The main work is as follows:(1)Aiming at the problem that the common spatial pattern cannot extract time and frequency domain features,a feature extraction algorithm combining the intrinsic time scale decomposition LCD and the common spatial pattern based on motor imagery EEG is proposed.First of all,select a part of EEG signal channels to be decomposed by LCD to obtain a series of intrinsic scale components(ISC).Then,the ISC components and EEG signals are integrated and processed to extract the spatial and time-frequency features.Finally,the features of the four-class of motor imagery EEG signals were fused and classified.The results show that the proposed method improves the recognition rate of EEG signals by fully extracting the time-frequency and spatial information from the motor imagery signals.(2)Aiming at the redundancy problem and high dimensionality of feature vector in the process of feature extraction,combining the firefly algorithm and learning automata,this paper proposed an optimized feature selection algorithm,which can effectively optimize the feature vector.Compared with the genetic algorithm(GA)and adaptive weight particle swarm optimization algorithm(APSO),the results indicate the proposed algorithm can effectively reduce the feature dimension,eliminate redundant features,and improve the classification accuracy of EEG signals at the same time,which indicates the good performance of our method.(3)Designed and constructed a motor rehabilitation training system based on motor imagery,and realized the motor rehabilitation training for the disabled based on the classification of four-class motor imagery.This system receives data collected from UE-16 B EEG amplifier by socket,and calls the MATLAB for filtering,feature extraction,feature selection and classification of EEG data.Finally,the results will be feedback to the motor rehabilitation training system and the subject.On this basis,the system designed and completed the training modules with and without feedback and the BCI game.It can promote users to exercise and finish the rehabilitation training smoothly.