Brain computer interface approach using sensor covariance matrix with forced whitening

Abstract

In this paper, we present a novel motor imagery classification method in electroencephalography (EEG)-based Brain-Computer Interfaces (BCIs) using forced whitened sample covariance matrices as features. The proposed method performs a constant-forcing to the weaker sources of covariance matrices before a whitening process to prevent amplifications of noise sources which have small power relative to class relevant sources. Experimental results show the improved accuracy in comparison with a classification without forced whitening process.