An adaptive calibration and control method for EMG control of a robot arm
Patients with chronic neuromuscular conditions such as spinal cord injury are left with reduced motor function in much of their body, which makes it difficult to interact with their environment. Robot arms controlled with residual muscle signals measured with electromyography give such individuals the chance to complete activities of daily living on their own. While such technologies are helpful in returning a degree of autonomy to these individuals, these systems must still be tuned to the specific user to accommodate the unique nature of their injury. This thesis presents a novel method of EMG control together with a calibration procedure to adapt the controls to the needs of the user without resorting to manual tuning by an expert. This method allows a user to freely move the end-effector of a robot arm, but does not rely on the health or usability of any particular muscle. Results from a pilot study evaluating the performance of this method show that in comparison to previously used methods, the smoothness of motion is improved, speed of movement is comparable, but accuracy is lower. The increased smoothness of motion suggests that greater control over movement speed and direction is possible with further practice and familiarization. While tested in a two-dimensional environment, the method has the potential for a generalization to control an arbitrary number of dimensions with an arbitrary number of muscles, providing a single solution to a wide variety of applications and users.