[Signal processing in electromyography: a review]

Automatic analysis of electromyography (EMG), signals, first operated in 1950 with analogic machines, steeply expanded from 1980 when fast computers and worthwhile programs became available. On-line measurement of response area and latency, averaging of low amplitude waves, fast sorting of motor unit potential shape parameters, computation of the "jitter" between two muscle fibers, turns/amplitude and spectral analysis of interferential pattern records, are some examples of programs currently offered in modern EMG machines. Other techniques are still reserved for research purposes: scanning EMG, decomposition of nerve and muscle compound potentials, measurement of the threshold and firing rate of motor units, trace analysis using tracking models. Finally, the credit for artificial intelligence systems (knowledge based systems, fuzzy logic, neuronal networks) is still not clearly stated.