Activity recognition in long-term electromyograms.

Long-term electromyography with portable recorders allows the study of muscle activity in the natural environment to investigate whether muscle overuse or oral habits may contribute to initiate or perpetuate a myoarthropathy of the masticatory system. At present, little is known about the behaviour of masticatory muscles over long time periods. The aim of this preliminary study was to define parameters which allow the automatic recognition of different types of oral activities from the electromyogram form. A programme with functional and parafunctional activities was performed by four volunteers (total of 333 functional and 82 parafunctional recordings). Electromyograms of the masseter and temporal muscles were recorded by means of a self-developed portable intelligent solid state recorder. Signal mean level (mw) and dynamics (dw) within a sliding window were determined. Temporal muscle recordings allowed better discrimination between function and simulated parafunction. The average of mw for clenching was 51.4 +/- 3.5% and for tooth grinding 21.2 +/- 2% of the peak electromyogram value at maximum voluntary clench. At 0.96 s sliding window duration, clenching, tooth grinding and chewing signals had maximum separation, using dw/mw as parameters (average: 0.16 +/- 0.01 for clenching, 0.39 +/- 0.01 for teeth grinding and 0.88 +/- 0.01 for chewing).