Detection of P300 components using the Wiener filter for BCI-based spellers

We present a novel method to reduce noise in EEG to acquire better event-related potentials (ERPs) and therefore detect P300 components more accurately. Our method employs the Wiener filter to remove noise from a desired P300 waveform. The desired signal for the Wiener filter is constructed by using the average P300 waveform observed in the training data. Accuracy of detecting P300 components increases when we use the ERPs filtered by the Wiener filter, compared to other cases using conventional filters. This suggests that the performance of the P300 speller may be improved when the Wiener filter is used to reduce noise in EEG.