Extraction of Neural Control Commands using Myoelectric Pattern Recognition: a Novel Application in adults with cerebral Palsy

This study investigates an electromyogram (EMG)-based neural interface toward hand rehabilitation for patients with cerebral palsy (CP). Forty-eight channels of surface EMG signals were recorded from the forearm of eight adult subjects with CP, while they tried to perform six different hand grasp patterns. A series of myoelectric pattern recognition analyses were performed to identify the movement intention of each subject with different EMG feature sets and classifiers. Our results indicate that across all subjects high accuracies (average overall classification accuracy > 98%) can be achieved in classification of six different hand movements, suggesting that there is substantial motor control information contained in paretic muscles of the CP subjects. Furthermore, with a feature selection analysis, it was found that a small number of ranked EMG features can maintain high classification accuracies comparable to those obtained using all the EMG features (average overall classification accuracy > 96% with 16 selected EMG features). The findings of the study suggest that myoelectric pattern recognition may be a useful control strategy for promoting hand rehabilitation in CP patients.