Brain-Computer Interface speller using hybrid P300 and motor imagery signals

In this paper we propose a fast brain computer interface speller based on electroencephalography (EEG). The slow performance of conventional BCI spellers is overcome by combining the fast motor evoked potentials (MEPs) with the accuracy of P300 event related potentials. The .mu rhythms associated with motor imagery are extracted using morlet wavalet based time-frequency analysis. Selected features were subsequently classified using minimum Mahalanobis distance. A hybrid MEP-P300 algorithm incorporating text prediction was proposed and experiments were conducted to gauge its accuracy and speed. Results show significantly faster performance when compared with conventional P300 spellers while comparable, but reduced accuracy was also noted.