PERIODIC SPATIAL FILTER FOR SINGLE TRIAL CLASSIFICATION OF EVENT RELATED BRAIN ACTIVITY

摘要

Because of the small amplitudes of event related potentials(ERPs), they are usually hidden in electroencephalogram(EEG) recordings. This is particularly a problem when analyzingsingle-trial data. A spatial filtering method for P300detection in oddball paradigm is proposed in this paperwhich is based on the assumption that brain responses tothe same stimulus look the same (or at least do not changesignificantly over trials). Therefore, the sequence generatedby concatenating all the responses to the same typeof stimulus has a hidden periodicity. Enhancing the periodicstructure of this sequence, a transformation is foundto project the data into a lower dimensional subspace. Experimentsshow that even with a small subspace of the projecteddata, the classification performance in single-trialP300 detection is still high.