Study On The Classification Algorithm Of EEG Signals Based On Filter Bank And Sparse Representation

A brain-computer interface(BCI)is a communication technology that does not rely on the brain's peripheral nerves and muscles to transmit directly information to the human brain and external devices or to control external devices.BCI can provide People with diseases such as brainstem stroke and spinal cord injury who cannot communicate normally with the outside world.In the BCI system,the electroencephalography(EEG)signal is the expression form of the brain nerve signal and is the basis of signal processing in the system.However,the EEG signal may change greatly in different people at different times.Therefore,the processing of EEG signals becomes particularly critical.The process of EEG signal processing includes data preprocessing,feature extraction,and feature classification.Each part affects the classification performance of the entire system..Therefore,in order to improve the classification performance of BCIs based on motor imagination,this paper proposes to design an optimized filter bank to decompose EEG signals into subband signals with different frequency bands,and then to select the optimal subbands.The optimal subband combining Sparse Representation(SR)is a new algorithm for feature extraction.This method improves the three parts of data preprocessing,feature extraction and feature classification.It can select the optimal frequency band according to different people and different time and use it for classification,and finally improve the classification performance of the entire system.This paper uses two data sets of the third and fourth BCI competitions to test the comparison of this method with the Combining Wideband(8-30Hz)CSP with Fisher Discriminant Analysis(FDA)and Combining Broadband CSP with Sparse Representation methods.The experimental results show that the method in this paper,the combination of filter bank and sparse representation can select the optimal frequency band based on the experimental data of different subjects at different times,and obtain the best classification performance,verifying the feasibility and effectiveness of the algorithm.