Functional mapping of multiple mechanomyographic signals to hand kinematics

Current methods of prosthesis control are overly simplistic, using two opposing electromyographic control signals to control the prosthetic hand, normally with no finger control. Recent research has demonstrated that mechanomyographic (MMG) signals can do the same. This experiment investigates the possibility of fusing multiple MMG signals to create a control scheme for prostheses that provides control over groups of fingers. Concurrent recordings of MMG activity and finger motions were made during several finger movements. The recordings were used to determine the functional mapping that exists between MMG signals and finger motions, and to implement a basic classification system. Two hyperplanes were able to separate thumb from finger flexion with 87% accuracy and pinkie from middle three finger flexion with 82% accuracy. Overall accuracy was 76.2%. Further improvements can be made over this linear classifier, and are worth further study. The results indicate that MMG may be used to enhance accuracy in remote manipulation applications.