A Study on Signal Processing Algorithms for Visually Evoked P300 Potential in EEG
|School||Guangdong University of Technology|
|Course||Control Theory and Control Engineering|
|Keywords||Brain - machine interfaces P300 Wavelet Analysis Support Vector Machine Character classification|
Brain - machine interfaces (brain-computer interface, BCI) is to establish a direct exchange of information between the human brain and a computer or other electronic device and control channel, it is widely used in the field of medical care, rehabilitation, entertainment and even military prospects. Is a common visual evoked P300 EEG for brain - machine interface system. EEG traditional P300 EEG signal processing methods as a complex, non-stationary random signal superimposed times and low classification accuracy. P300 EEG signal processing technology, the main contents are as follows: Firstly, traditional P300 EEG processing algorithms for EEG data processing. Butterworth digital filter for low-pass filtering, using the optimal weighted average superposition algorithm to remove the random noise. Pick relatively standard P300 waveform as a match with the template, after pretreatment EEG waveforms with standard templates do the analysis of the correlation number of the size of the character classification, the classification accuracy rate of 64.52%, based on their mutual relations. Character classification accuracy rate in the original template matching process a P300 EEG classification on the basis of its improvements, the highest classification accuracy rate of 77.42%. Traditional template matching P300 EEG classification using only 1 lead data in 64-lead data, character classification accuracy rate is low. To take full advantage of multi-lead data analysis to further improve the character classification accuracy, this paper designed a use of the wavelet transform and support vector machine combined P300 EEG feature extraction and classification. The first to use the the optimal weighted average superimposed algorithm to increase the signal-to-noise ratio of the P300 EEG. Wavelet decomposition of to extract P300 EEG characteristics, at the same time play the role of the low-pass filtering, data compression, P300 identification will affect the larger guide 10 associated number as the character classification input, can effectively extract P300 EEG characteristics. P300 EEG data classification using support vector machine method to samples with fewer training can achieve relatively high P300 EEG classification accuracy. RBF kernel function to calculate the optimal parameter training model for cross-validation, MATLAB LIBSVM tool simulation the character prediction correct rate of 83.87%, significantly better than traditional P300 EEG processing methods. In this paper, the wavelet transform and support vector machine P300 EEG feature extraction and classification, compared with the traditional template matching algorithm can effectively improve the classification accuracy of P300 EEG based visual evoked potential P300 real-time BCI system has laid a foundation.