A new method of analyzing the surface electromyogram (SEMG) has been developed, in which each SEMG is characterized by a single "representative surface motor unit action potential" (RSMUAP). This method is based upon modelling the SEMG as the output of a linear time-invariant system that has an impulse response equal to the RSMUAP. The RSMUAP can then be recovered from the SEMG using a technique called bicepstral deconvolution. Bicepstral deconvolution is a homomorphic method of separating signals that is based upon the bispectrum. Using the bispectrum, rather than the power spectrum is required for a non-minimum phase estimation of the RSMUAP. The bispectrum has the added benefit of being zero for Gaussian signals, making it well suited for analyzing noisy data. Simulations indicate that this new method of SEMG analysis should be effective for detecting the progression of myopathy