P300 feature extraction for visual evoked EEG based on wavelet transform

Abstract

It is a crucial issue to accurately and quickly extract the feature of visual evoked potentials in the brain-computer interface technology. Based on the non-stationary and nonlinearity of the electroencephalogram (EEG) signal, a method of wavelet analysis is adopted to extract P300 feature from visual evoked EEG. Firstly, the imperative pretreatment for EEG signals is performed. Secondly, the approximate and detail coefficients are gotten by decomposing EEG signals for two layers using wavelet transform. Finally, the approximate coefficients of the second layer are reconstructed to extract P300 feature. The results have shown that the method can accurately extract the P300 feature for visual evoked EEG, and simultaneously, obtain time-frequency information which traditional methods can not do. Therefore, wavelet transform provides an effective method to feature extraction for EEG mental tasks. © 2011 Springer-Verlag.