A threshold selection criterion based on the number of runs for the detection of bursts in EMG signals

摘要

In this work, a method to calculate the threshold value in the detection of activity phases of electromyographic signals is introduced. The criterion used by the method is based on the number of runs produced by the threshold process. The novelty of this technique lies in that the threshold value is theoretically determined employing the statistical properties of the number of runs. This value is adjusted then to the empirical distribution of the data in the signal. We present some experiments with synthetic and experimental signals to demonstrate the performance of the proposed method. Moreover, we compare it against other recent methods available in the literature. Our results show a superior performance with respect to other recent approaches. (C) 2019 Elsevier Ltd. All rights reserved.