Study And Implementation Of Brain Computer Interface System Based On Visual Evoked EEG

The technology of Brain-computer interface can build a new "brain-computer-device" channel,which does not rely on traditional peripheral nerve and muscle tissue.The patients with severe disability can use the brain-computer interface to control an external device or communicate with the environment.Brain-computer interface technology involved in neuroscience,cognitive science,control science,and other disciplines.With the development of computer technology and the brain science,brain-computer interface technology has become a hot area of research at home and abroad.This paper mainly studies the brain-computer interface system based on visual evoked EEG signals,the steady-state visual evoked potential and p300 are two common visual evoked potentials in the research of brain-computer interface.In this paper,the key technologies of the steady-state visual evoked potential based brain-computer interface and p300 based brain-computer interface are studied,the work mainly include the design of visual stimulation,EEG signal acquisition,EEG signal processing,and online control system design.For the SSVEP has typical feature of frequency,we respectively employ wavelet decomposition and reconstruction,Fast Fourier Tansform and Canonical Correlation Analysis in SSVEP preprocessing,SSVEP feature extraction and SSVEP pattern classification.Offline experiment has collected eight subjects' EEG data under the four types of tasks,then the wavelet combining canonical correlation analysis is used to classify the four different types of SSVEP and the average accruracy rate is above 85%.In this paper,the steady-state visual evoked potential based online brain-computer interface system is implemented.The subjects can control the NAO robot move to different directions through choosing one of the four visual stimulation on the computer screen with different flashing frequency.Eight subjects use the brain computer interface system to control the movement direction of the NAO robot,the average accuracy 87.5% were obtained finally.The p300 has both time-domain characteristics and certain frequency distribution characteristics,so the independent component analysis combining band-pass filter,the superposition averaging combined with wavelet packet transform and the support vector machine(SVM)pattern classification method is respectively used for p300 EEG preprocessing,feature extraction and pattern classification.Six subjects' p300 EEG data have been collected in the offline experiment.The result shows that the pattern classification accuracy of six tasks achieved above 80%.This research firstly focuses on the offline data analysis of steady-state visual evoked potential and p300,the corresponding signal processing algorithms have been choosed based on the different characteristics of different potential and it has achieved a high accuracy.In addition,this paper has implemented an online brain-computer interface system,which is easy to use and has high accuracy and information transmission rate.So this reseach has certain significance of promoting brain-computer interface technology to the practical application.