Time-frequency based classification of the myoelectric signal: static vs. dynamic contractions

This work represents ongoing investigation in pattern recognition for myoelectric control. It is shown that four channels of myoelectric data greatly improve the classification accuracy, as compared to two channels. Also, it is demonstrated that the steady-state myoelectric signal may be classified with greater accuracy than the transient signal. The exceptionally accurate performance of the four channel system using steady-state data suggests that a robust online classifier may be constructed, which produces class decisions on a continuous stream of data. This would represent a more natural and efficient means of myoelectric control than one based on discrete, transient bursts of activity.