Implementation And Studies On SSVEP-Based Brain-Computer Interface

Brain-Computer Interface(BCI), which as a new kind of human-computer interaction isbecoming a hot topic of brain research, has open up a new pathway for information exchangebetween the human brain and the outside world, and it has a great potential for applications inthe field of rehabilitation engineering, high-risk operations, psychological cognitive, et al.More and more scientific groups participate in the research of BCI and have completed manydifferent BCI systems, but for the time being, BCI still stay in the stage of research inlaboratory, and faces with great challenge before practical application. The topic of this paperis a BCI based on the Steady-State Visual Evoked Potential (SSVEP), which has a highresearch value because of the advantages of noninvasive signal recording, little trainingrequired for use, and high information transfer rate.On the basis of access to extensive literature, this paper first introduces the basicdefinition of BCI, and summarizes the status and significance of the BCI. Then according tothe characteristics of the SSVEP, and reference to the basic design methods and ideas aboutBCI, achieve a BCI experimental system based on SSVEP, including the realization of thevisual stimulation, the selection of stimulation parameters and the design of experimentalmethods.Perform a large number of experiments on the BCI experimental system and then offlineanalysis the experimental data. The analysis uses two frequency extraction algorithms, whichinclude Fast Fourier Transform, multi-resolution Mallat wavelet and AR model. The twoalgorithms can effectively extract the frequency characteristics of the SSVEP. To analyze andcompare the accuracy of the two algorithms for data processing, then select the algorithmbased on FFT to carry on follow up study.The applications of BCI system require very high data processing accuracy. This paperproposes a generally applicable electrode-combination selection method to improve theaccuracy of the data processing. According to the model of the electrode signal, divide thesignal on occipital electrode into effective signal and noise signal based on IndependentComponent Analysis (ICA) theory. Then propose several electrode-combination methods, andselect the electrodes whose effective signal is significant to participate in theelectrode-combination selection. Finally determine the suitable electrode-combination methodfor each subject by comparing the data processing accuracy of different method, and provethat the electrode-combination selection can improve the data processing accuracy.On the basis of the study above, this paper complete the online experiment of the BCI system and achieve a simple character input function. So it achieves the whole BCI systembased on SSVEP and has laid a foundation for its application.