Signal analysis of the electromyogram

The author has postulated that the relevant information of the human neuromuscular system during voluntary muscle contractions is contained in the surface electromyogram, or EMG. This information is important in diagnosis, modelling and rehabilitation, and consists of two parts. The first is the action potential waveform of motor units, and the second the random firing pattern of several such units. This hypothesis is borne out by the existing state of art in electrophysiological analysis and modelling. Rigorous statistical analysis of the EMG of three muscles of volunteers showed gross similarities in variance, statistical bandwidth and spectral moments, which suggested that a common physical phenomenon was involved. A biphasic waveform model of the action potential was formulated, and its estimated parameters obtained from spectral-moment functions showed strong similarities for different subjects, as well as for all three muscle groups. On the basis of these preliminary analyses, and using digital filtering techniques, the EMG spectra were decomposed, with the expectation of obtaining the action-potential waveform and the firing-patterns. This was indeed found to be so, the spectrum off the model closely agreeing with that obtained from spectral decomposition of the real signal. Moreover, a realistic average waveform of the action-potential was obtained, in addition to estimates of the number of such waves in each data segment. A set of analytical results and statistics of similarity-parameters are presented, considered to be the first of their kind anywhere to be basedon the decoded surface electromyogram alone.