Novel formulation of a double threshold algorithm for the estimation of muscle activation intervals designed for variable SNR environments.

The aim of this work is the development of an improved formulation of the double threshold algorithm for sEMG onset-offset detection presented by Bonato and co-workers. The original formulation, which keeps the threshold fixed, suffers from performance degradation whenever the SNR changes during the analysis. The novel approach is designed to be adaptive to SNR changes in either burst or inter-burst zones of sEMG signals recorded in static and dynamic conditions. The detection parameters (i.e. detection and false alarm probabilities) are updated on the basis of an on-line estimation of the SNR. The proposed formulation has been assessed on both simulated and real sEMG data. For constant SNR the performance of the original formulation is confirmed (for SNR > 8 dB, bias and standard deviation less than 10 and 15 ms, respectively; detection percentage higher than 95%), while the novel implementation performs better with time-varying SNR (for SNR varying in the range 10-25 dB the standard approach detection percentage decreases at 50%). Detection on signals recorded during isometric contractions at different force levels confirms the performance on simulated signals (StD = 134 ms; FP = 22%, and StD = 42 ms; FP = 2%, respectively for standard and novel implementation calculated as average on five experimental trials). The pseudo real-time detection allowed by this formulation can be profitably exploited by biofeedback applications based on myoelectric information.