Impact of the time segment analysis for P300 detection with spatial filtering

Abstract

A Brain-Computer Interface (BCI) allows the direct communication between humans and computers by analyzing brain activity. The oddball paradigm allows detecting event-related potentials (ERPs), like the P300 wave, on targets selected by the user. While this paradigm provides the location of the P300 wave in the signal, its exact location remains a hypothesis and depends on the subject. This paper deals with the choice of the time segment for the signal analysis and its impact on the classification. A method for selecting the relevant part of the signal that contains the P300 wave is proposed. First, spatial filters are estimated for enhancing the signal. Second, a part of the enhanced P300 wave is selected based on its magnitude. This selection aims at providing an optimal start for the time window representing the P300 wave. Three window lengths are compared. We show that a window length of 500ms provides on average the best results, but the optimal window length should be set individually. The proposed technique has been validated on data recorded on 20 healthy subjects.