Algorithm for identification of motor unit action potentials based on wavelet transform and neural networks

Nowadays, it is common to identify some neuromuscular disorders from the myoelectric signals (MES). Often, these disorders are reflected in the basic components of the MES, the motor unit action potentials (MUAP). This work presents an approach for the decomposition of intramuscular MES in its essential MUAPs, through analysis (wavelet transform) and classification (neural networks) tools. Decomposition aims to obtain the largest number of MUAP and its features. The wavelet transform was used to identify the MUAPs; after, an artificial neural network was implemented as a first approach of classification; and finally, a second sorting was carried out through the firing rate. As a result, in a record were identified more than 100 MUAP and these were grouped into three classes with 2 subclasses each one. Finally firing rates and average errors for each group were obtained.