Classification of the mechanomyogram: its potential as a multifunction access pathway

Although the mechanomyogram (MMG) has been demonstrated as a viable representation of muscle activity, its potential as a multifunction (>2) control signal has not yet been investigated. This study investigates the discriminability of multiple hand motions using multichannel forearm MMG. With nine able-bodied participants, MMG signals from six sites could be differentiated among eight classes of forearm muscle activity with a mean accuracy of 93 +/- 9% using 15 features selected by a genetic algorithm and classified by a linear discriminant analysis classifier. These results suggest that, with additional research, MMG may indeed become a usable control signal for multifunction access devices.