Research On EEG Signal Processing Based On Steady-state Visual Evoked Potential

Brain-computer interface(BCI)technology is a new type of human-computer interaction,which can convert EEG signals into instructions that can be recognized by the computer to control the external equipment.The BCI systems based on steady-state visual evoked potential(SSVEP)have attracted wide attention due to their high operability and stability,however,the accuracy of target recognition still has a large room for improvement.For SSVEP signal processing,the existing algorithms fail to make full use of its signal features and individual differences.Therefore,this paper focuses on SSVEP preprocessing,feature enhancement and signal classification,and proposes optimization algorithms based on the existing algorithms,so as to improve the performance of SSVEP-BCI system.The main content of this paper includes the following parts:(1)In terms of SSVEP signal preprocessing,this paper selects 9 channels in the occipital region of brain as the optimal electrode combination according to the response mechanism,signal characteristics and single channel experimental analysis of SSVEP.Moreover,filter processing and independent component analysis(ICA)are used to reduce the noise of the original signal,remove the environmental noise and the common biological artifact interference,and thus improve the signal-to-noise ratio of the signal.(2)In the aspect of SSVEP feature enhancement,a feature enhancement algorithm SCMEMD based on multivariate empirical mode decomposition(MEMD)is proposed,and a new construction mode of auxiliary signal is designed.The addition of auxiliary signal alleviates the effect of mode aliasing and redundant noise in the process of signal decomposition,thus improving the identification degree of signal features.The frequency features of the signals which decomposed,screened and reconstructed by this algorithm are significantly enhanced,it is conducive to the classification of SSVEP in the future.(3)In the aspect of SSVEP signal classification,this paper focuses on the optimization and improvement of canonical correlation analysis(CCA)algorithm.Because the traditional CCA algorithm ignores the harmonic information and the signal difference between different subjects in SSVEP,in this paper,we propose DS-CCA algorithm and SC-MEMD-CCA algorithm based on in-depth utilization of harmonic information,as well as the SSVEP classification algorithm DIT-CCA based on multi-band individual signal template.DS-CCA and SC-MEMD-CCA make full use of the harmonic information of SSVEP from two different angles of reference signal and EEG signal respectively.DIT-CCA divides the frequency band of individual signal templates to make them have individual differences,at the same time,the harmonic information in the signal template is deeply utilized.The experimental results show that these three algorithms have a significant effect on improving the accuracy of SSVEP classification,and their accuracy is 5.83%,6.83%and 9%higher than that of the traditional CCA algorithm respectively.Finally,we fuse the features of DS-CCA algorithm and DIT-CCA algorithm,the frequency recognition accuracy is up to 70.17%,which is 12.17%higher than the traditional CCA algorithm.