Recently, covariance matrices have been shown to be interesting features for signal classification and object detection. In this paper, we review and compare the existing kernels on covariance matrices and explore their use for EEG classification in Brain-Computer Interfaces (BCI). This study addresses both experimental and theoretical aspects of the problem. Beside the apparent complexity of the kernels, we show that this approach simplifies the whole BCI system. Finally, we empirically demonstrate that this simpler approach obtains state-of-the-art results on the BCI competition IV dataset 2a.
Download Full PDF Version (Non-Commercial Use)