Discrimination of intended motions for prosthetic hands using nonstationary EMG

The present paper proposes a multiple neural network to determine the movement intended by an amputee from electromyogram (EMG) signals. Most previous approaches to the discrimination of movements using EMG signals have required measurement data with a relative long period exceeding 200 ms. Our approach is possible to identify the amputee's intended movement based on EMG signals of 70 ms long in an initial rise zone. Experiments with four subjects and four electrode locations demonstrated that our proposal determines six forearm movements at a discrimination rate exceeding than 90%.