Volume conductor simulation and neural network decomposition of surface EMG

A muscle of finite length distributions in a finite limb is used to derive surface MUAPs by volume conductor theory. Active fibers within a motor unit are lumped to a line source located at the motor unit center, coordinates of which are generated from a uniform distribution. Surface MUAP trains are formed by scheduled firing frequency with random perturbations. H surface EMG is formulated by superposing surface MUAP trains with a uniformly distributed initial phases. ART2 Neural Network is used to decompose the simulated surface EMG. Successful decomposition up to a maximum of five MUAPs is demonstrated for either simulated surface EMG or recorded surface EMG in weak voluntary isometric contraction