Clinical applications of myoelectric signal processing by neural network and spectral analysis

Recent works have shown that the myoelectric signal processing allows to obtain many informations about brain activity during muscle contractions. The knowledge of these informations can be used in order to implement useful rehabilitation methods for neuropatologic patients. Our goal is to give some techniques ables to evaluate muscle functions/dysfunctions in clinical applications by means of neural network and spectral analysis. In this paper we show that during the contraction of postural muscles like pectorals a cerebral low-frequency common drive has found. For this purpose, the simultaneous activities of both pectoral and both first interosseous muscles are recording by surface electromyography. Subsequently, an independent component analysis neural network is performed in order to remove artifacts. The discovery of the common drive in the brain was performed by means of the coherence analysis of myoelectrical signal recorded, based on Welch method spectrum estimate. The obtained results are in agreement with clinical studies.