Classification of upper limb motions in stroke using high density surface EMG

Myoelectric pattern recognition techniques have been developed to infer user's intention of performing different functional movements, which can be used to provide volitional control of assisted devices for people with disabilities. The pattern recognition based myoelectric control systems have rarely been designed for stroke survivors. Aiming at developing such a system for stroke rehabilitation, this study assessed the myoelectric control information remained in the affected limb of stroke survivors using high density surface electromyogram (EMG) recording and pattern recognition techniques. The experimental results from 3 stroke subjects indicate that high accuracies (92.42% ± 5.51%) can be achieved in classification of 20 different intended movements of the affected limb. This study confirms that substantial motor control command can be extracted from paretic muscles of stroke survivors, potentially facilitating their rehabilitation.