EEG Classification Algorithm Based On Steady State Visually Evoked Potential

Brain-Computer Interface(BCI)is an interaction system connecting brain and computer.BCI is one of the most popular research area in current brain science,which combines with information science.From the ways of brain signal acquisition,BCIs can be divided into invasive and noninvasive BCI systems,in which the Steady State Visually Evoked Potentials(SSVEP)-and Motor Imagery(MI)-based noninvasive BCIs are the most popular ones in current studies and applications.EEG is a weak and noisy electric signal,mixed with a large number of artifacts which causes interference to the processing and identification for EEG signals.Using EEG preprocessing methods can effectively remove artifacts,noises,and improve EEG signal-noise ratio,so as to provide more accurate EEG for its feature extraction and classification.In this work,we proposed a modified EEG processing approach named CS-CSP to improve its classification,which adopts genetic algorithm-based channel selection algorithm to reduce dimensions of feature vectors achieved by CSP algorithm.The experimental results show that the Chebyshev type II filter is superior over conventional preprocessing methods and that CS-CSP greatly improves the recognition accuracy compared with the solo CSP.A template-matching approach is proposed in this thesis to enhance the performance of SSVEP-based BCI,combining with Multivariate Synchronization Index(MSI)and Independent Component Analysis(ICA)-based spatial filter for SSVEPs frequency recognition.To show the superiority of MSI-ICA,we compared with CCA-,FBCCA-and ICA-based frequency recognition methods on SSVEP benchmark datasets provided by Tsinghua University.The experimental results demonstrate that the MSI-ICA-based method outperforms others with respect to classification accuracy and ITR.To provide useful EEG signals for BCI,this work focuses on EEG preprocessing.This work introduces the research,design and implementation steps of SSVEP stimulus in detail and explains the parameter selection for various stimuli.The Pygame library in Python is used to implement the SSVEP stimulus presentation program.In the end,we introduced the design of standard EEG signal acquisition system and stimulation paradigm.