Advances in processing of surface myoelectric signals: Part 2.

The paper focuses on the analysis of myoelectric evoked potentials and their progressive scaling, as well as morphological changes using orthogonal basis functions with essentially finite time support; and the compression of the information content using principal component analysis. The application of the Hermite-Rodriguez and the associated Hermite functions is discussed as a means to provide compact information about the shape of the M-wave or of the power spectral density function of either voluntary or electrically elicited myoelectric signals; a means to estimate scaling factors; and a means to describe and classify nonstationarities. The principal component analysis shows the possibility of a compression ratio of at least 10:1 in the storage of M-wave sequences. The paper also describes three methods for the estimation of delay between similar signals, and therefore for estimation of conduction velocity. They are based on normalised integrals. Fourier transform matching and matching in the time domain. In particular conditions they provide different results for the same pair of signals. The concept of delay and the performance of these methods are reviewed and discussed. The paper is not exhaustive. It has the main objective of making the reader aware of the wealth of methods available for nonstationary myoelectric signal analysis and conduction velocity estimation, and of the need to use them with knowledge of their respective advantages, disadvantages, peculiarities and limitations.