The Research Of Evoked EEG Feature By Audio-visual Stimulus

The information of pictures, speeches and texts related to audio-visual perception plays an important role in the social activities. It can be not only directly perceived and comprehended by human, but also processed through the computer. However, the ability of computer processing is lower than the human’s ability. And the processing efficiency can not satisfy the development of nowadays society. Utilizing the mankind’s cognitive mechanism and research achievement, we can establish a new calculating model to improve the comprehension ability and processing efficiency of computer. And it not only will boost the rapid development of information science, but also provide a power for brain cognitive science. Currently, the research has obtained great achievement in alone visual and auditory stimulus channel of brain. However, the researches are exploring for the perceptual and cognitive function of brain’s cross-modality, EEG feature related to audio-visual cognitive activity, brain area fixed position and brain function network analysis, as well as the identifying technique of audio-visual fusion information. Especially, those key problems, including the paradigm design of audio-visual stimulus, the precision and efficiency for feature extraction and recognition of EEG signal, will still not be resolved fully.Considering above problems, we researched the test paradigm of synchronous audio-visual stimulus, the method to the feature extraction and recognition for evoked EEG. The completed work is shown as follows:(1) We designed the test paradigm about the visual, auditory and the combined audio-visual stimulus. Then, the EEG signal was acquired by EEG amplifier from visual evoked, auditory evoked and audio-visual synchronously evoked pattern.(2) In order to eliminate the measuring noises, interference and artifacts, the original EEG data were preprocessed using the methods of common average reference, digital bandpass filter and independent component analyze. (3) We performed the wavelet transform, bispectrum analysis and wavelet-bispectrum algorithm to extract the EEG feature using the data provided by BCI Competition Ⅲ. The simulation results have shown that the three methods are effective for the visual evoked feature extraction. And then, the wavelet transform was performed to extract EEG feature for visual, auditory and audio-visual synchronous evoked EEG. Finally, ERP of brain was compared in the way of the three evoked stimulus.(4) The method to multi-channel EEG data combined with PNN was presented for the recognition of audio-visual evoked EEG features. Then, we used the method to classify features for the data of BCI Competition Ⅲ and the test data of paradigm in this paper. The results have shown that this method is better in classification accuracy, compared with other methods.The research achievements in this paper could be applied to the brain’s cognitive science and BCI system. And the research can lay the foundation for the neural information processing, meanwhile, has the double values to science and the application.