Real-Time Control of an Exoskeleton Hand Robot with Myoelectric Pattern Recognition

Robot-Assisted training provides an effective approach to neurological injury rehabilitation. To meet the challenge of hand rehabilitation after neurological injuries, this study presents an advanced myoelectric pattern recognition scheme for real-Time intention-driven control of a hand exoskeleton. The developed scheme detects and recognizes user's intention of six different hand motions using four channels of surface electromyography (EMG) signals acquired from the forearm and hand muscles, and then drives the exoskeleton to assist the user accomplish the intended motion. The system was tested with eight neurologically intact subjects and two individuals with spinal cord injury (SCI). The overall control accuracy was 98.1 ± 4.9% for the neurologically intact subjects and 90.0 ± 13.9% for the SCI subjects. The total lag of the system was approximately 250ms including data acquisition, transmission and processing. One SCI subject also participated in training sessions in his second and third visits. Both the control accuracy and efficiency tended to improve. These results show great potential for applying the advanced myoelectric pattern recognition control of the wearable robotic hand system toward improving hand function after neurological injuries.