Extraction And Application Of Visual Evoked Potentials

Visual evoked potentials(VEPs)are electrical responses in the occipital region of the cerebral cortex when subjects receive visual stimulation.At present,the research on VEP mainly has two applications.Firstly,in clinical applications the VEP can reflect whether lesions exist in the visual pathways,and is also important in the diagnoses for white matter malnutrition,spinal cord cerebellar disorders,Parkinson,tumor,etc.Secondly,because of the stability,VEP can be used as information transmission media which realizes human-computer interaction,i.e.the Brain-Computer Interface(BCI)research based on VEP,with advantages of training free and multiple command sets.In reality,the single stimulus generated VEP signal is very weak,often submerged in the strong spontaneous EEG and noise,and so it is difficult to directly distinguish it from the recorded waveform.The existing methods to extract VEP have their individul drawbacks.The most commonly used superimposed averaging method requires too many stimuli.Blind signal separation(BSS)method needs large amount of calculation or the prevalent machine learning method requires a lot of Data training.However,in the clinical diagnosis or BCI applications,the VEP extraction should be real-time,i.e.the VEP should be quickly extracted from fewer stimuli.Furthermore,from the BCI research perspective,the application of VEP not only concerns the extraction problem,but also considers the specific task for experimental design,signal processing and other issues.In view of the above problems,this thesis focuses on research of VEP from two aspects: the extraction problem and the application problem.The main work of this thesis is as follows:In one aspect,from the perspective of VEP extraction,a method was proposed to extract VEP signals based on improved Wiener filter in case of a small number of visual stimuli.The improved Wiener filter expresses the VEP signal as a combination of the useful signal and the interfering signal orthogonal thereto.The defined useful signal can optimally characterize the main features of the VEP by means of orthogonal representation,such as the P100 component.The goal of this method is to maximize the signal-to-inference-noise ratio between the useful signal and the interfering signal and noise,which is more efficient than the conventional Wiener filter to extract useful information from the VEP.However,there was still a problem that the cross-correlation matrix estimation was insufficient in the case of improved Wiener filtering,and hence we effectively estimated the cross-correlation matrix by extending the collected VEP data into periodic signals and gived the improved Wiener filtering method for VEP extraction in this case.Through experiments on multiple subjects,we showed that this method can extract VEP quickly,and proved its reliability referring to the latency of P100.In the other aspect,from the perspective of application of VEP,we proposed a chromatic transient visual evoked potential(CTVEP)experiment system based on pseudo random sequence modulation color vision stimulation,and designed a BCI system that enables the subjects to choose a specific location on the screen through "ideas".In the designed system,different chromatic con-central rings exhibited in different positions on the screen simultaneously,with on/off pattern each specified by one distinct sequence from multiple pseudo random Gold sequences.For the received data from occipital region,the BCI system gived a demodulation method of CTVEP based on the orthogonality among Gold sequences,and ultimately the subject focusing area on the screen was given by a matched filtering method.Through experiments on multiple subjects,the identification accuracy of the proposed system was given,and the reliability of the system for translating the ideas of the subjects was verified.