Suboptimal sensor subset evaluation in a P300 brain-computer interface

A Brain-Computer Interface (BCI) is a specific type of human-computer interface that enables the direct communication between human and computers by analyzing brain activity. Oddball paradigms are used in BCI to generate event-related potentials (ERPs), like the P300 wave, on targets selected by the user. This paper deals with the choice of a reduced set of sensors for the P300 speller. A low number of sensors allows decreasing the time for preparing the subject, the cost of a BCI and the P300 classifier performance. A new algorithm to select relevant sensors is proposed, it is based on the backward elimination with a cost function related to the signal to signal-plus-noise ratio. This cost function offers better performance and avoids further mining evaluations related to the P300 recognition rate or the character recognition rate of the speller. The proposed method is tested on data recorded on 20 subjects.