This paper presents a new approach to the decomposition of electromyographi
c (EMG) signals, EMG signals consist of a superposition of delayed finite-d
uration waveforms that carry the information about the firing of different
muscle fiber groups. The new approach is based on a communication technical
interpretation of the EMG signal, The source is modeled as a signaling sys
tem with interssymbol-interference, which encodes a well-defined sparse inf
ormation sequence. This point of view allows a maximum-likelihood (ML) as w
ell as a maximum aposteriori (MAP) estimation of the underlying firing patt
ern to be made. The high accuracy attainable with the proposed method is il
lustrated both with measured and artificially generated EMG signals.