Adaptive filters for muscle response suppression

To be able to use the voluntary EMG-signal from an electrically stimulated muscle as control signal for FES-applications, it is necessary to eliminate the muscle response evoked by the stimulation. The muscle response is a non-stationary signal, therefore a set of linear adaptive prediction filters are proposed, based on the observation that the shape of the muscle responses only exhibits moderate changes during a time window of up to 300 ms. The filters are derived and compared with a conventional fixed comb filter on both simulated and real data. For variations in amplitude of the muscle responses the performance of the adaptive filters are independent of the filter length, whereas for variations in the shape the performance is increased with the filter length. Using the adaptive filters it is possible to obtain a signal-to-noise ratio, which enables the EMG from a partly paralysed muscle to be used as control signal for stimulation of the same muscle