Multiple tangent space projection for motor imagery EEG classification

Abstract Due to its non-invasiveness and easiness implement, EEG signals decoding are in base of most based brain computer interfaces (BCI) studies. Given the non-stationary nature these signals, a preprocessing phase is needed. An interesting idea perform use spatial covariance matrices. In last years, matrices was extensively used electroencephalography (EEG) signal processing filtering for Motor imagery (MI) BCI. Spatial lie Riemannian manifold Symmetric Positive-Definite (SPD) matrices, therefore, geometry attracting lot attention showing be simple, robust, providing good performance. This paper explores enhancing information provided classifier by combination different projections from their native space multiple class-depending tangent spaces. We demonstrate that this new approach provides significant improvement model accuracy.