Classification Performance Of Different Motor Unit Action Potential Feature Space Representations: A Simulation Study
Daniel William Stashuk, R.K. Naphan · 2005
The template matching classification performance and apparent separation matrices of several different motor unit action potential (MUAP) feature space representations have been studied. Feature spaces considered included: 1) high resolution time sample, 2) Nyquist resolution time sample, 3) Nyquist resolution frequency coefficient and 4) Nyquist resolution power spectrum coefficient representations of the MUAPs. Representations I), 2) and 3) all had essentially equal performance, while power domain representations performed poorly. The MUAPs studied were extracted from simulated, point source electrode, myoelectric (ME) signals and therefore were contaminated with background ME activity and could be optimally aligned.