Research On SSVEP Classification Algorithm Based On Correlation Analysis

Brain-Computer Interface(BCI)aims to establish a direct communication channel between the brain and the external environment,thus allowing individuals with brain injury to regain the ability to communicate with the outside world.The BCI has a variety of stimulation paradigms,and the Steady-State Visual Evoked Potential(SSVEP)based on BCI has been widely studied and applied in this field because of its short training time,easy to use and high information transfer rate.In recent years,the design of stimulus paradigm systems and signal detection and classification algorithms in the field of SSVEP have been greatly improved.Among these metheds,Canonical Correlation Analysis(CCA)is one of the mainstream frequency recognition methods in brain-computer interfaces based on steady-state visual evoked potentials.Since the requirements for classification accuracy and information transfer rate are update,various improved CCA algorithms have been proposed by researchers in this field.In the recent study,the Latent Common Source Extraction(LCSE)framework algorithms based on Generalized Canonical Correlation Analysis(GCCA)have achieved excellent results on the open source SSVEP benchmark dataset from Tsinghua University,and they significantly outperforms other methods in terms of both classification accuracy and information transfer rate.However,the classification accuracies of these methods need to be improved under short time windows,and the following research work is conducted in this paper for sovling this problem.The sparse GCCA algorithm is investigated in the framework of LCSE based on GCCA algorithm and it is applied to SSVEP signal detection classification.Also,the Extended Generalized Canonical Correlation Analysis(Exd-GCCA)method is used to obtain more correlation coefficient features to process the problem of low classification accuracy in short time windows.In addition to,by combining the method of correlation coefficient feature fusion with Task-Related Component Analysis(TRCA)algorithm and Correlated Component Analysis(CORCA)algorithm,we propose the Extended Task-Related Component Analysis(Exd-TRCA)algorithm and the Extended Correlation Component Analysis(Exd-CORCA)algorithm are proposed.Experimental results show that our Exd-TRCA and Exd-CORCA classification results are better than the basic framework algorithms,and Exd-GCCA method outperforms the LCSE algorithm in terms of classification accuracy and information transfer rate.Therefore,these methods have great potential for implementing high-performance BCI systems based on SSVEP.