Phase Synchronization Features and Common Spatial Patterns for the Classification of Motor Imagery Data

Owen Falzon and Kenneth P. Camilleri

Keywords

braincomputer interfaces, common spatial patterns, electroencephalography, phase synchronization

Abstract

The interaction of distributed neuronal assemblies in the brain is thought to play a significant role during cognitive tasks. Various measures have been developed to quantify these interactions and a number of research groups have evaluated the usefulness of these measures in braincomputer interface (BCI) setups. In BCI systems these interaction measures have mainly been used to complement features representing other characteristics in the data. In this study we consider the combination of phase synchronization measures and features obtained from the method of common spatial patterns (CSP). The latter technique, although widely adopted for BCIs, has never been considered in conjunction with phase synchronization measures. For the phase synchronization features we make use of the phase synchronization based CSP (P-CSP) method. We thus carry out a number of tests on motor imagery EEG data to determine whether the combination of features from the two techniques can improve BCI performance. The results from these experiments indicate that the combination of P-CSP features with features from the CSP technique does not yield any significant improvement in classification. Nevertheless, consistent phase synchronization patterns are obtained across subjects, indicating a common underlying synchronization process that emerges during the motor imagery tasks.

Important Links:



Go Back