Electromyogram signals contain information for predicting muscle force that can be used in human-machine interaction and medical applications such as the control of prosthetic hands. Different methodsexist for estimating the SEMG–force relation. However, muscle dynamic variations during voluntary con-tractions due to fatigue have been neglected in the identification stage. This would make the models notapplicable to normal working conditions. We developed a model based on Laguerre expansion technique,LET, to identify the dynamic SEMG–force relation and investigate the presence of fatigue through kernelanalysis. Our proposed data acquisition protocol was used to induce fatigue in the muscles involved inthe act of grasping, hence enabling us to study the effects of muscle fatigue. The results of LET in com-parison with fast orthogonal search and parallel cascade identification, which were able to accuratelyidentify the desired dynamics, represent an improvement of 15% and 3.8% in prediction fitness, respec-tively. Moreover, by extracting median frequency (MDF) of the recorded SEMG signals and tracking itschanges over time, the existence of muscle fatigue was studied. The results showed that fatigue had animpact on the Brachioradialis muscle. The first and second order kernels of the LET illustrated variationsin the time and frequency domains similar to that of MDF for the Brachioradialis muscle correspondingto the fatigue generation process. Employing the proposed model the dynamics of SEMG–force relationcan be predicted and its variations due to muscle fatigue can also be investigated.

Keywords