A neural network approach to decomposing surface EMG signals

A neural-network-based approach to decomposing surface EMG signals into a multitude of single muscle-fiber action potentials (SFAP) is presented. This decomposition is done to yield the signal forms of the SFAPs and to allow for localizing the particular SFAPs relative to the recording (surface) electrodes. The goal of this approach is to allow the physician or medical researcher to observe the waveforms of the electrical activity and to localize SFAPs below the surface of the skin in a noninvasive manner and without discomforting the patient, as is important for several neurological diagnostic purposes. The approach uses a Hopfield neural network in combination with a Kohonen network for discrimination and a conventional computer correlation algorithm