A New BCI Classification Method based on EEG Sparse Representation

Abstract

Motor imagery based Brain Computer Interface (BCI) systems provide a new communication and control channel between the human and an external device with only imagination of limbs movements. Because Electroencephalogram (EEG) signals are very noisy and nonstationary, powerful classification methods are needed. We propose a new classification method based on sparse representation of EEG signals and ell-1 minimization. This method requires a well constructed dictionary. We show very high classification accuracy can be obtained by using our method. Moreover, our method shows improved accuracy over a well known LDA classification method.