System Design And Algorithm Research Of SSVEP-based Muti-output BCI

A brain-computer interface(BCI)provides a communication pathway which independent of peripheral nerves and muscles for individuals with severe motor neuron diseases.BCI systems based on steady-state visual evoked potentials(SSVEPs)have high classification accuracy,information transfer rate,and signal-to-noise ratio,giving them high research and application value.However,existing research on SSVEP-based BCI still has several limitations.Firstly,the current SSVEP coding methods need lots of stimulus frequency to encode output commands,which makes frequency detection more difficult.Secondly,the existing decoding algorithms cannot meet the design requirements of BCI system with high classification accuracy and plug-and-play.Therefore,this study attempts to optimize the system design and classification algorithm of SSVEP-based multi-output BCI.The obtained results are as follows:1)In terms of SSVEP coding method,we proposed a novel stimulus coding method named mode based time frequency coding(MTFC)method.The MTFC method adopts the mode based classification decision mechanism to overcome the problem of single trial accuracy decline caused by the cumulative multiplication of the accuracy of each time segment in the traditional time-frequency coding method.Meanwhile,the MTFC method retains the less frequency coding characteristics of the time-frequency coding method,which can code 48 commands using only four frequencies.In this study,we designed a SSVEP-based BCI spelling system which containing 48 characters using MTFC method.The experimental results show that the classification accuracy of BCI system designed by MTFC method is 6% higher than that of the BCI system designed by the traditional Multiple Frequencies Sequential Coding(MFSC)method.2)In terms of SSVEP decoding algorithm,we proposed FBCCA+STBF algorithm which combining the advantages of filter bank canonical correlation analysis(FBCCA)and spatiotemporal beamforming(STBF).FBCCA+STBF algorithm inherits the training-free characteristics of FBCCA,so the dedicated training process is not necessary.In addition,the classification accuracy of FBCCA+STBF algorithm can maintain the same level as the trained STBF algorithm.The obtained results show that FBCCA+STBF method can achieve 78.7%modified accuracy,which is 4.4% higher than FBCCA method.In conclusion,the proposed MTFC method and FBCCA+STBF algorithm can improve the performance of SSVEP based multi-output BCI system from two aspects of encoding and decoding.The SSVEP based BCI spelling system designed by MTFC and FBCCA+STBF algorithm has the characteristics of high performance and plug and play.