A NEURAL-NETWORK APPROACH TO ELECTROMYOGRAPHIC SIGNAL-PROCESSING FOR A MOTOR CONTROL TASK

A hybrid modeling structure composed of a one degree of freedom comput ational musculoskeletal model and a feedforward multi-layer perceptron neural network was used to effectively map electromyography (EMG) fro m a human exercise trial to muscle activations in a physiologically fe asible and accurate fashion. Several configurations of the complete hy brid system were used to map four muscle surface EMGs from a ballistic elbow flexion to normalized muscle activations, estimated individual muscle forces and torque about the joint. The net joint torque was use d to train the neural portion of the hybrid system to minimize kinemat ic error. The model allowed the estimation of the nonobservable parame ters: normalized muscle activations and forces which was used to penal ize the learning system. With these parameters in the learning equatio n, our system produced muscle activations consistent with the classic triphasic response present in ballistic movements.