| Brain-Computer Interface(BCI)is a technology that directly interacts with the external environment by acquiring and analyzing brain waves,independent of peripheral nerve and muscle tissues.BCI systems have been widely applied in medical diagnosis,disability assistance,smart homes,entertainment and many other fields.Recently,Steady State Motion Visual Evoked Potential(SSMVEP)-based BCI,which uses mild stimulus paradigm to acquire electroencephalograph(EEG)signals,has received extensive attention in application.This dissertation designs and implements an SSMVEP stimulus paradigm and verifies its feasibility and practicability in the BCI system.On the other hand,with the deepening of BCI research,the problem of "BCI Illiteracy" has gradually become prominent.In order to improve the classification accuracy of EEG signals of different subjects,and to understand and ameliorate the problem of “BCI Illiteracy”,based on BCI system,this dissertation makes an in-depth analysis and classification research of SSMVEP EEG signals.The main works are as follows:1)This dissertation designs and implements the ring-shaped motion checkerboard paradigm,constructs the visual stimulation interface,and verifies the stimulation interface through online experiments.The verification results show that the designed stimulation interface can successfully induce the corresponding SSMVEP signals.Based on the stimulation interface,EEG acquisition experiments are carried out to acquire SSMVEP signals from two groups of subjects,namely“EEG literate” and “EEG illiterate”.2)In this dissertation,the influence of individual differences on SSMVEP classification tasks is verified by analyzing the EEG signals of two groups of subjects.This dissertation develops the multivariate weighted recurrent network methodology and constructs a multivariate weighted recurrent brain network based on SSMVEP signals,and characterizes the brain network by the weighted local efficiency and the weighted clustering coefficient,thus exploring the mechanism and deep-seated reasons of brain cognitive behavior of poor performance of EEG illiterate subjects in BCI application from the perspective of complex networks.This method provides a new way for SSMVEP signal analysis and the exploration of "BCI illiteracy".3)In this dissertation,a multi-domain neural network model is constructed by combining convolutional neural network(CNN)and long-term and short-term memory network(LSTM),which realizes the extraction and fusion of various features in the time domain,frequency domain and space domain of SSMVEP signals,and is applied to the classification and identification research of SSMVEP signals.The results show that compared with traditional methods,the classification accuracy of SSMVEP signals for all subjects,including EEG illiterate subjects,can be improved,and the robustness and universality of SSMVEP-based BCI system are enhanced to a certain extent. |
