Phase Synchronization Indices for Classification of Action Intention Understanding Based on EEG Signals.

The classification of action intention understanding based on EEG signals is very important for human-robot and social interaction studies. In order to classify the action intention understanding brain signals efficiently, we first use three kinds of phase synchronization indices, phase locking value (PLV), phase lag index (PLI) and weight phase lag index (WPLI), to construct functional connectivity matrices in multiple micro time windows, and then extract the sum of significant edge values of each time window matrix as the classification feature, finally apply support vector machine (SVM) classifier to implement action intention understanding data classification task. Classification result shows that new method performs well on three datasets (alpha, beta and fusion frequency bands), and brain network statistical analysis demonstrates that many significant edges appear on the alpha frequency band. We conclude that the phase synchronization indices are extremely useful for the classification task, the sum of significant edge values is an effective classification feature, and the action intention understanding closely correlates with the alpha frequency band.