Research On Brain-Computer Interface Based On Visual Stimulus

Brain-computer interface(BCI)is a kind of control technology that does not depend on human peripheral nerves and muscles,but establishes a direct path between human’s brains and computers or other external devices.People can directly operate the external devices by the activities in their brain,which provides not only the possibility for disabled people who have lost their ability to communicate with the external world,but also an alternative way for healthy individuals.The brain-computer interfaces based on visual stimuli,with the advantages of the high information transfer rates,the easy operation,and the short training time,haves been widely studied.In this paper,we mainly focus on brain-computer interfaces based on Steady-State Visual Evoked Potential(SSVEP)and Rapid Series Visual Presentation(RSVP),and investigate the design patterns of experiments,the processing pipline of EEG signals and the setup of real time systems.Firstly,we improved the traditional algorithm used in the SSVEP-BCI EEG recognition tasks based on existing researchs,by proposing an algorithm called Filter Bank Likelihood Ratio Test(FBLRT),which should make up for the disadvantages of the Likelihood Ratio Test(LRT)algorithm not taking full used of the harmonic components of SSVEP signals.Secondly,we developd a word spelling system based on the SSVEP-BCI using the sampled sinusoidal stimulation.Data synchronization is realized by sending flag bits to parallel ports.Then,we proposed two different types of filter banks,and collected and analyzed offline data by the SSVEP-BCI word spelling system,and selected the optimal filter bank parameters according to the results.Finally,we compared the recognition accuracy and information transferring rate of the LRT,the Filter Bank Canonical Correlation Analysis(FBCCA)and the FBLRT proposed in this paper by both offline and online experiments respectively.The results demonstrated that the performance of the FBLRT is significantly better than the other two algorithms.Setting the stimulation time to be 1s,the average recognition accuracy rate achieved 84.96%,he ITR 160.01 bits/min.The highest ITR was 208.35 bits/min.The performances of SSVEP-BCI system have been greatly improved,which promoted the practical application of SSVEP-BCI.Besides,we investigated image EEG recognition tasks via a RSVP-BCI.First,we clipped the high-resolution radar satellite(Synthetic Aperture Radar,SAR)images into 512*512 samll pictures,designed a experimental RSVP paradigm,and built a target recognition system based on RSVP-BCI.Secondly,a processing flows based on wavelet analysis were introduced.We analyzed the single-trial EEG evoked by the RSVP paradigm,and compared the performance of Linear Discriminant Analysis(LDA)and Support Vector Machine(SVM)through off-line experients.Meanwhile,the method with the best overall performance was selected for online experiment.Finally,the wavelet-based processing method was tested online.The average true positive rate and the false positive rate is 84.38% and 14.11% respectively,and the Area Under Curve(AUC)can reach above 0.9,which comfirming the feasibility of the system we designed.