Onset Detection in Surface Electromyographic Signals: A Systematic Comparison of Methods

Abstract

Various methods to determine the onset of the electromyographic activity which occurs in response to a stimulus have been discussed in the literature over the last decade. Due to the stochastic characteristic of the surface electromyogram (SEMG), onset detection is a challenging task, especially in weak SEMG responses. The performance of the onset detection methods were tested, mostly by comparing their automated onset estimations to the manually determined onsets found by well-trained SEMG examiners. But a systematic comparison between methods, which reveals the benefits and the drawbacks of each method compared to the other ones and shows the specific dependence of the detection accuracy on signal parameters, is still lacking. In this paper, several classical threshold-based approaches as well as some statistically optimized algorithms were tested on large samples of simulated SEMG data with well-known signal parameters. Rating between methods is performed by comparing their performance to that of a statistically optimal maximum likelihood estimator which serves as reference method. In addition, performance was evaluated on real SEMG data obtained in a reaction time experiment. Results indicate that detection behavior strongly depends on SEMG parameters, such as onset rise time, signal-to-noise ratio or background activity level. It is shown that some of the threshold-based signal-power-estimation procedures are very sensitive to signal parameters, whereas statistically optimized algorithms are generally more robust

Trunk muscle activation during golf swing: Baseline and threshold

“…The onset burst was detected at 22%Pmax with method A and 33% with method B and the agreement with VI was low (R -ICC = 0.365; L -ICC = 0.214), but method A showed a better correlation. A higher signal-to-noise ratio leads to a better resemblance between the methods, and the reverse situation otherwise (Staude et al, 2001). Both automatic methods detected the onset peak in ES, which is a timing pattern in all the subjects characterized by intensely turning off/turn on the muscular activation.…”

“…The difficulty in comparing studies is related to the different algorithms criteria, which compromises the reproducibility of the results (Morey-Klapsing et al, 2004;Jöllenbeck, 2000). This is particularly evident for threshold algorithms (Staude et al, 2001).…”

“…The first factor refers to the difficulty in comparing different studies, due to a low reproducibility method (Allison, 2003;Morey-Klapsing et al, 2004;Staude et al, 2001), when there is some diversity among algorithms (Hug, 2011). Another factor that was considered refers to the tendency of preferably testing algorithms in maximum voluntary contractions or in monoarticulate actions (e.g.…”

“…However, the lack of a goldstandard measurement used to validate the algorithms, leads to visual inspection being used to assess the precision of threshold algorithms. Algorithms Detection can be classified into threshold algorithms (Van Boxtel et al, 1993;Hodges and Bui, 1996;Jöllenbeck, 2000;Allison, 2003) and as statistically optimized algorithms (Micera et al, 1998;Staude et al, 2001), as maximum likelihood.…”

“…The precision with which a certain algorithm detects the onset is influenced by the background activity level, signal-to-noise ratio activity (Hodges and Bui, 1996;Staude et al, 2001), and onset rate of signal amplitude (Allison, 2003). Hug (2011) states that threshold algorithms vary in 1, 2, and 3 SD or between 15% and 25% of the activity's maximum peak.…”