Automatic recognition of motor unit potentials

The study of myoelectric activity by deep electromyography usually does not permit recording the activity of a single motor unit. In favorable cases some motors can be recognized by amplitude criteria using threshold systems. Otherwise, the traces must be analyzed ″manually″ , measuring the pulse spacings with a ruler, a long and exacting task. For such cases it seems desirable to commit this analysis to a computer. It is then necessary to establish an automatic recognition system discriminating by waveform rather than by amplitude. A system for the automatic recognition of motor unit potentials is presented; it makes possible the retrieval of the dates of discharge of a single unit in a graph in which several units are collated simultaneously. This system operates by autoinstruction, i. e. the system operates when a single sample of the unit sought is fed into the computer in plain language. The percentage error is low where the unit sought is of characteristic shape, time-stable, and in quiet surroundings.