Research On BCI Online Neurofeedback Based On Motor Imagery EEG

Electroencephalograph(EEG)of motor imagery(MI)based brain computer interface(BCI)can be used to replace the original or rebuild the loss output of the central nervous system.The original intention and technology requirement of neural feedback are similar to BCI.EEG is the most convenient neurophysiological signal in nerve feedback.The research of EEG signal processing method can accelerate the development of neural feedback technology.This article mainly takes the motor imagery EEG as the research object,which have studied the preprocessing,feature extraction and classification of less channel EEG,and preliminary designed and tested the online neural feedback system.This paper mainly studies from the following aspects,and has obtained the corresponding research results.(1)Preliminary designed an EEG music player,users can control the music player by EEG of different limb motor imagery,the player can also play the corresponding types of music automatically by recording and processing the EEG from users in different emotion,to adjust the mood of users,the design enhance the intelligent degree of music player and adaptive feedback and adjustment function of user emotion based on EEG.(2)In view of the problem of inadequate input channels of online less-channel EEG using independent component analysis directly and automatic recognition of the artifacts,this paper proposed a method of time delay data extract window to expand the dimensions of less-channel EEG,and then consider the multi-channel EEG after dimension expansion as the input of fastica directly,then calculate kurtosis values of each independent component and identify,remove the EOG components automatically.(3)Study the influence of phase synchronization combined with autoregressive model in less-channel motor imagery EEG to the classification accuracy.Hilbert transform is used to calculate the phase locking value(PLV)of EEG,and combined with the 6 order AR model parameters to form multi-dimensional feature vectors.The parameter optimized SVM classifier is applied to classify the multi-dimensional features of EEG,so as to improve the classification accuracy of the classifier.By considering the single AR model power spectrum estimation and the phase synchronization feature method as comparison experiment,it proved that the combination of AR model power spectrum estimation and phase synchronization have the advantage of extract the motor imagery EEG feature.(4)Preliminary built and tested the online motor imagery neural feedback system.The left and right hand motor imagery combined with teeth is used to control the Tetris game.In the game,the rotation of the Tetris and the state of the topographic map on the interface are used as feedback to enable the subjects to concentrate their attention and adjust their motor imagery strategies in real time.The experimental results proved that the visualization online training system can enhance the motor imagery and neural function plasticity through neural feedback,and hopefully lay a foundation for the clinical application of nerve feedback in the treatment of neurological disorders.