Optimal autoregressive orders for myopathic electromyograms.

This paper aims to describe the optimal autoregressive order of varying-length electromyograms for myopathic subjects. Epochs of electromyography signals are modeled as outputs of autoregressive systems, for orders varying from 1 to 100. The optimal order to represent each epoch is chosen by the minimum description length criterion. Probability density functions are fitted to the histograms of the optimal orders. The lognormal function provides the best fitting, and its mean value varies linearly with the epoch length.