EEG Signal Analysis And Operation Platform Based On Visual Evoked Potential

Brain computer interface(BCI),as a medium for the exchange of information between the brain and the physical device,has received extensive attention from the fields of neuroscience,cognitive science,rehabilitation therapy and so on.Among of the BCI,steady state-visual evoked potentials(SSVEP)have been widely studied because of its high signal-to-noise ratio,short response time and short training time.However,the information transfer rate of BCI still cannot meet the requirement.Based on this,this dissertation conducts the research from two aspects,one is the LCD refresh frequency effect on the performance of SSVEP-BCI,the other is the study of signal recognition algorithm.This dissertation discusses the SSVEP potential and the underlying information of EEG,so as to improve the performance of SSVEP-BCI.Specific work as follows:(1)The effect of LCD refresh frequency on the performance of SSVEP is studied.At present,LCD is widely used in the SSVEP-BCI system as the stimulating device,but the parameter selection of refresh rate of the LCD is still a blank.In view of this problem,this dissertation investigates the effects of different refresh rates(60Hz,75 Hz,100Hz,120 Hz,144Hz)on the performance of SSVEP-BCI under the same conditions.The results show that the higher the refresh frequency is,the higher the information transmission rate of SSVEP-BCI system is in the short-term(no more than 2S)stimulation.(2)The conventional filter bank Canonical Correlation Analysis(Filter bank CCA,FBCCA)algorithm and the individual template training(it3-CCA)algorithm are improved in this dissertation,and we present a combined algorithm of two improved algorithms.In the conventional FBCCA algorithm,the power function is used as the formula to calculate the weight.This dissertation makes use of the downhill stationary characteristic of exponential function,and designs an improved FBCCA algorithm(exp-FBCCA algorithm).Offline experimental results show that the performance of exp-FBCCA is better than FBCCA,and the information transfer rate is 102.35bits/min when the stimulus time is 1.5s.In order to solve the problem that the feature measurement of it3-CCA is too little,a new feature measurement(it5-CCA),which is composed of spatial filter and training data set,is introduced to get the deep information of the signal.Then,in this dissertation,by jointing the advantages of exp-FBCCA and it5-CCA,a SSVEP-BCI system based on combined algorithm is established.Online and offline results show that the combined algorithm is superior to FBCCA and it3-CCA.When the stimulating time is 0.5s,the information transfer rate is 242.71bits/min,which meets the actual needs of high speed BCI system.(3)In the SSVEP-BCI system,Canonical Correlation Analysis(CCA)algorithm is the core of most of the signal recognition algorithms.Based on this premise,this dissertation compares the performance of the Likelihood Ratio Test algorithm(LRT)and the CCA algorithm in SSVEP-BCI.The online and off-line experiments show that the performance of LRT is better than that of CCA.Then,this dissertation discusses the feasibility of using LRT to replace CCA in the extended algorithms,which are based on canonical correlation analysis.(4)The BCI operating platform based on SSVEP potential is researched.The dissertation introduces the solution of thread paradigm programming,signal synchronous method and multi thread processing,then an online experiment based on this solution is carried out.The results show that the SSVEP-BCI system is stable and reliable.