Three-mode classification and study of AR pole variations of imaginary left and right hand movements

Abstract

Classification of left and right hand imaginary movements is a fundamental task in Brain Computer Interface (BCI) applications. In this paper we propose a three-mode classifier which can distinguish between no hand movement (background), imaginary left hand movement and imaginary right hand movement. This is distinct from the twomode classifier which is typically used in BCI. We show that in this three-mode environment, an EM initialized Kalman smoother gives much better classification scores, particularly for the background mode, than a randomly initialized smoother. A further contribution of the paper is a study on the variations of the autoregressive poles with imaginary hand movement. It is shown that during the movement period, the magnitude of the poles which correspond to the alpha and beta bands for the contralateral EEG signals, undergo significant changes. This correlates with the well established ERD phenomenon known to be present in this type of data.