Time-varying delay estimation with application to electromyography

Pathological or physiological state of the muscle can be assessed from the velocity of propagation of surface action potentials (conduction velocity - CV). The estimation of CV from surface electromyography (sEMG) implies an estimation of time delay between signals detected by two or more sensors along the muscle length. In this paper we investigate the possible use of a parameter estimation approach to follow changes of CV over time. The recursive least square algorithm was used. The error on estimation of CV was quantified in the case of Gaussian white noise (GWN) and band-limited signals. On this second type of signal, a decimation and a whitening filter were used to increase the robustness of the algorithm in case of additive noise. The results indicate that the frequency bandwidth substantially affects performance. The best performance was reached with GWN. For band-limited signals, the decimation processing followed by whitening substantially increased the quality of the estimation.