Binary Classification Of Hand Movement Directions From Eeg Using Wavelet Phase-Locking

Phase synchronies are often used to study relationships between different parts of the brain and to identify regions that interact in a coordinated manner for a certain task. In this paper, we propose a wavelet reconstruction and phase-locking-based feature extraction method to visualize and classify the direction-specific phase synchronies between Electroencephalogram (EEG) channel-pairs for hand movements in 4 directions using EEG data collected from 7 subjects performing right hand movements. We then study its discriminative ability by using statistical analysis and report the most informative, direction-specific channels and wavelet levels. Next, we show the discriminative performance of the proposed feature extraction method in the binary classification of 6 direction pairs. Subsequently, we use the Minimum Redundancy Maximum Relevance feature selection algorithm to select features which improved the classification accuracy of our proposed method by 4.39%. Thus, the results demonstrate the potential of proposed wavelet phase-locking method to extract movement direction related information from EEG.