BUILDING THE FOUNDATION OF CLINICAL PRACTICE OF EMG PATTERN RECOGNITION FOR PROSTHETIC ARM CONTROL

This project improves the function of upper limb prostheses by developing a reliable, robust, and clinically-viable prosthesis control system based on electromyography (EMG) pattern recognition (PR). Conventional prosthesis control (i.e. body-powered or proportional EMG control) is inadequate for multifunctional prostheses operation. Research in laboratory has shown that EMG PR enables transradial (TR) amputees or above-elbow amputees with targeted muscle reinnervation surgeries to control multiple degrees of freedom of a prosthesis intuitively and efficiently. Unfortunately, no commercially available prosthetic arms use EMG PR control scheme due to several challenges for clinical practice, including high computational complexity, lack of wearability, poor robustness, and need for frequent recalibrations. The objective of this project is to develop new technologies and engineering solutions that resolve the difficulties in current EMG PR-based prosthesis control, advancing its adoption in practice. The design incorporates: (1) an optimized EMG PR algorithm for accurate, reliable, and responsive user intent recognition; (2) novel sensor fault detectors, system recovery technologies, and spatial filtering approaches to ensure the robustness of the sensor interface and EMG PR system; (3) a new wearable and user-friendly calibration interface integrated with a prosthesis-guided calibration program; and (4) embedded implementation of advanced control algorithms specifically tailored to the hardware structure for fast and accurate algorithm execution with power efficiency.