A statistical method for the measurement of muscle activation intervals from surface myoelectric signal during gait

The aim of this work is to present an original double-threshold detector of muscle activation, specifically developed for gait analysis. This detector operates on the raw myoelectric signal and, hence, it does not require any envelope detection. Its performances are fixed by the values of three parameters, namely, false-alarm probability (Pfa), detection probability, and time resolution. Double-threshold detectors are preferable to single-threshold ones because, for a fixed value of the Pfa, they yield higher detection probability; furthermore, they allow the user to select the couple false alarm-detection probability with a higher degree of freedom, thus, adapting the performances of the detector to the characteristics of the myoelectric signal of interest and of the experimental situation. In this paper, first we derive the detection algorithm and describe different strategies for selecting its parameters, then we present the performances of the proposed procedure evaluated by means of computer simulations, and finally we report an example of application to myoelectric signals recorded during gait. The characterization of the proposed double-threshold detector demonstrates that, in most practical situations, the bias of the estimates of the on-off transitions is smaller than 10 ms, the standard deviation may be kept lower than 15 ms, and the percentage of erroneous patterns is below 5%. These results show that this detection approach is satisfactory in research applications as well as in the clinical practice.