Designing Phase-Sensitive Common Spatial Pattern Filter to Improve Brain-Computer Interfacing.

This paper addresses an interesting problem to model common spatial pattern (CSP) using an objective function employed to segregate EEG signals for a given cognitive task into two classes. The novelty of the present research is to include phase information of the EEG signal along with the amplitude for differentiating class boundaries. Two modified CSP algorithms are proposed in this paper. The first one introduces the composite effect of amplitude and phase angle of the EEG signal in CSP formulation and is solved using Lagrange's multiplier method taking phase information of EEG into account. In the second approach, a novel CSP algorithm is proposed in this paper which has the efficacy of handling the non-linearities hidden in the brain signal, here EEG. Experiments undertaken confirm that the proposed phase-sensitive CSP yields the best performance than their non-phase sensitive counterparts by a large margin with respect to classification accuracy.