Analyzing fatigue in dynamic exercise through electromyography signals and similarity metrics

Abstract

Electromyography (EMG) is a technique that registers the electrical activity from the muscle; one application for this kind of signal is muscle fatigue analysis, which can be defined as the clue of losing muscle strength during physical activities. These activities can involve isometric or dynamic contractions, whereas, for the last one, there is no consensus about the more precise technique for muscle fatigue analysis. This work aims to compare three different analysis techniques to study muscle fatigue using EMG signals recorded during a pedaling exercise named Wingate test (dynamic contractions): median frequency (MDF), correlation coefficient (CC), and mutual information (MI). The techniques were applied in the frequency domain and, for CC and MI, were observed the similarity between the end and beginning of the signal using two kinds of references: fixed and variable. If the difference between these parts is identified using the references, it successfully identifies muscle fatigue. Comparing the results, CC and MI got better results than MDF, which reached 70.2% of the expected result, which we consider as indicative of fatigue. The results with CC and MI were more satisfactory: for fixed reference, CC got 77.2% of indicative of fatigue and 94.7% with MI. Variable reference got the best results: CC got approximately 98.2% and MI 96.5%, respectively.