Research On EEG Preprocessing Algorithm In Brain-computer Interaction Of Motor Imagery

Brain computer interface(BCI)is a new way of human-computer interaction,the motor imagery electroencephalogram(EEG)based BCI is one of the most commonly used BCI system,at present,the performance of the BCI system is not up to the requirements of products and practical,especially the need for further research and improvement of the EEG preprocessing method.EEG are easily contaminated by various physiological and non-physiological artifacts.Traditional or single preprocessing method can not meet the performance requirements of BCI system.This paper focuses research on the preprocessing method of motor imagery EEG,which put forward a new kind of EEG preprocessing method,quantitative evaluating the performance of this method by simulated EEG data,and conduct the EEG microstate research to off-line motor imagery EEG data,then,do a online verification and application to the preprocessing method,in the last,designing an efficient preprocessing platform of motor imagery EEG on the basis of above research.The main research contents are as follows:(1)Proposed a motor imagery EEG preprocessing method which combined discrete wavelet transform(DWT),fast independent component analysis(fastica)with clustering analysis(DWICA+ clustering analysis).firstly,converting the less-channel EEG into multi-channel by DWT,and then direct consider the multichannel signal as the input of fastica,calculating the feature of each independent components in the field of time,spectrum and correlation between the sequence,and then use hierarchical clustering algorithm to cluster each independent component,automatically identify and remove artifact components,finally obtained the clean EEG signal.(2)Testing the preprocessing method of EEMD-ICA,DWT-ICA and DWICA combined with cluster analysis by simulated EEG,quantitatively evaluating the performance of these three methods by calculating the preprocessing time consumption,EEG signal to noise ratio and root mean square error respectively.Finally,the EEG microstate of the off-line motor imagery EEG data is studied,which lays a foundation for the application of microstate to the feature extraction of motion imagery EEG.(3)Do a online verification to the preprocessing method of DWICA combinedwith clustering analysis by combinating the feature extraction and classification method,extracting the instantaneous energy of subjects EEG in left and right hand motor imagery EEG during the experimental process,those features are classified by parameters optimized SVM,SVM and LDA classifier,the actual control results of 6subjects proved that the proposed preprocessing method of this paper is effectiveness in online application and the parameters optimized SVM classifier can further improve the classification accuracy.(4)Combined with the traditional EEG preprocessing method and the above research,from improving the system design speed and the algorithm transplantability point of view,preliminary designing the MFC platform of motor imagery EEG preprocessing,which laid a foundation for the productization and algorithm transplantation of BCI system based on motor imagery EEG.