Efficient parameterization of MUAP signal for identification of MU and volume conductor characteristics using neural networks.

The motor unit action potential (MUAPs) shapes depend on the anatomy and the physiology of the contracted muscle. The aim of this work is the identification of some characteristics of the motor unit (MU) and the volume conductor, namely the MU depth, the innervation zone width and the thickness of fat and skin layers based on MUAP signal parameters. The relationship between these characteristics and MUAP parameters are non-linear and complex. Thus, the use of the neural networks approach becomes an efficient tool to put in evidence this relationship. We have used the similarity and the homogeneity of the parameter criterions to choose which parameters are appropriate for the extraction. Two identification systems are presented and compared, a global system and a separate one. In order to evaluate the performance of each system, we have tested them using several simulated MUAP signals corrupted with additive Gaussian noise at different signal to noise ratios (SNR). A new test is introduced in which the electrode radius, the bar electrode dimensions and inclination angles for the detection system, fixed during the training process, are changed.