EMG-based Control For a Feeding Support Robot Using a Probabilistic Neural Network

This paper proposes a new manipulator control system to support the performance of eating tasks for people with severe physical disabilities, such as those with paralysis caused by cervical spine injuries. The system consists of an electromyogram (EMG) classification part, a manipulator control part and a graphical feedback display. It classifies the user's intended motions from EMG signals measured using a probabilistic neural network (PNN), and controls a robot manipulator in line with the results. Multiple subject motions can be accurately estimated based on learning of the user's EMG patterns using the PNN, thereby allowing operation of the manipulator as desired to perform eating tasks. To examine the performance of the proposed system, experiments were performed with five subjects, including one with paralysis from a cervical spine injury. The results demonstrated that the system could be used to accurately classify the subjects' EMG signals during motions, and that the unit could be easily controlled using EMG signals.