Classification of motor imagery tasks using phase synchronization analysis of EEG based on multivariate empirical mode decomposition

Abstract

Phase synchronization has been employed to study brain networks and connectivity patterns. The phase locking value (PLV) is one of the most effective measures widely used for phase synchronization analysis. We first calculate the PLVs of the pair-wise intrinsic mode functions (IMFs) based on multivariate empirical mode decomposition (MEMD) method. Next, the average PLV of the prominent pairs relative to the rest duration is adopted for the classification of motor imagery (MI) tasks. Comparative analysis with the EMD-based PLV method, the proposed method has a significant increase in feature separability for most subjects. This paper demonstrates that MEMD-based PLV method can provide an effective feature in the MI task classification and the potential for BCI applications.