A frequency domain Hampel filter for blind rejection of sinusoidal interference from electromyograms

A frequency domain filtering method is presented for the removal of power-line and other sinusoidal interference from electromyograms. The presence of such interference results in spikes in the complex-valued fast Fourier transform which can be considered to be outliers. The filter uses the outlier-robust Hampel identifier to detect spikes and replaces them with the median value of local values. The performance of the filter was tested using a set of simulated electromyograms and on experimental data. It was shown that the filter is more effective at rejecting interference than multiple notch filtering and spectral interpolation. The main advantages of using this method are that the frequencies at which the interference occurs need not be known a priori and that signals are not distorted in the absence of interference. In the case of single-frequency interference added to simulated and experimental data, errors in root-mean-squared amplitude and median frequency of <5% are reported.