Real-Time Detection of Myoelectric Hand Patterns for an Incomplete Spinal Cord Injured Subject

Abstract

Individuals with spinal cord injuries lose the ability to complete hand movements. Active orthosis based on myoelectric signals may provide a more intuitive control from the remaining muscles. Pattern recognition has been widely used to detect the intention to control assistant devices for rehabilitation, but little work has been extended to injured individuals. This work presents a proposal for real-time detection of hand movements based on myoelectric signals. A subject with incomplete spinal cord injury at the cervical level attempted to elicit flexion/extension fingers and resting while two-channel electromyographic signals were acquired. A classic on–off control was compared with different configurations of KNN, yielding classification performance up to 81.00% in real-time. The results showed the ability of the subject to performed contractions with repeated patterns for the control of low-cost active orthosis.