High-Density Myoelectric Pattern Recognition Toward Improved Stroke Rehabilitation

Myoelectric pattern-recognition techniques have been developed to infer user's intention of performing different functional movements. Thus electromyogram (EMG) can be used as control signals of assisted devices for people with disabilities. Pattern-recognition-based myoelectric control systems have rarely been designed for stroke survivors. Aiming at developing such a system for improved stroke rehabilitation, this study assessed detection of the affected limb's movement intention using high-density surface EMG recording and pattern-recognition techniques. Surface EMG signals comprised of 89 channels were recorded from 12 hemiparetic stroke subjects while they tried to perform 20 different arm, hand, and finger/thumb movements involving the affected limb. A series of pattern-recognition algorithms were implemented to identify the intended tasks of each stroke subject. High classification accuracies (96.1% +/- 4.3%) were achieved, indicating that substantial motor control information can be extracted from paretic muscles of stroke survivors. Such information may potentially facilitate improved stroke rehabilitation.