Real-time digital median frequency estimator for surface myoelectric signals

A prototype instrument is described that uses a digital signal processor to estimate the median frequency of the myoelectric signal in real time. A set of tenth-order digital low-pass filters is used to track the median frequency. An estimate of frequency is obtained once every 256 ms with a resolution of 1 Hz. The prototype implements the algorithm proposed by T.W. Hruczkowski and R.E. Gander (1984). The essential element of the algorithm is to compare the total power in the signal to the power in a low-frequency portion of the signal by means of estimating the variance in each of these. Test results confirm the operation of the prototype in real time. The algorithm is set up for use with surface myoelectric signals.<<ETX>>