Multivariate Fast Iterative Filtering Based Automated System for Grasp Motor Imagery Identification Using EEG Signals

One of the most crucial use of hands in daily life is grasping. Sometimes people with neuromuscular disorders become incapable of moving their hands. This article proposes a grasp motor imagery identification approach based on multivariate fast iterative filtering (MFIF). The proposed methodology involves the selection of relevant electroencephalogram (EEG) channels based on the neurophysiology of the brain. The selected EEG channels have been decomposed into five components using MFIF. Information potential based features are extracted from the decomposed EEG components. The extracted features are smoothed using a moving average filter. The smoothed features are classified using the k-nearest neighbors classifier. The cross-subject classification accuracy, precision, and F1-score of 98.25%, 98.31%, and 98.24%, respectively, is obtained. While the average classification accuracy, precision and F1-score for multiple subjects is 98.43%, 98.62%, and 98.41%, respectively. The proposed methodology can be used for the development of a low cost EEG based grasp identification system.