P300 Feature Extraction Based on Parametric Model and FastICA Algorithm

Abstract

A method based on AR model and Fast ICA algorithm for P300 feature extracting is presented. In the study, the visual evoked signal is obtained via the alternate pictures. Then, principal component analysis (PCA) is used for reducing the dimension of EEG signal, independent component analysis (ICA) is used for removing EOG artifact. And AR model is constructed for filtrating the spontaneous EEG. Finally, a coherence average is used to extract P300 in real-time. The results have shown that this method can perform effectively to extract P300 feature independently to any prior information and avoid the subject’s visual fatigue caused by long time visual evoking. It can be applied on online BCI system.