Shermin Shahbazi

In this paper, we propose a method to classify Electroencephalogram (EEG) signals for Brain Computer Interface (BCI) application on the manifold of Spatial Covariance Matrices (SCMs). We choose Riemannanian framework for analysis and projection to tangent spaces for transferring to vector spaces. The number of clusters and the basepoints of tangent spaces are computed automatically using a modified version of Angle Constraint Path (ACP) cluatering. One main advantage of ACP clustering is that it can determine the optimal number of clusters automatically a . In each tangent plane we classify projected SCMs using a SVM classifier, and finally we aggregate the prediction of different classifiers using an ensemble method .For experimental evaluation, we used dataset IIa from BCI competition IV and unbelievably accuracy values illustrate that our method outperform the other methods in classification.

Riemannian manifold, linearization, electroencephalogram signal, brain-computer interface, classification