Muscle Fatigue Assessment During Cycle Ergometer Exercise Using Principal Component Analysis of Electromyogram Power Spectra

Abstract

During muscle fatigue analysis some standard indexes are calculated from the surface electromyogram (EMG) as root mean square value (RMS), mean (Fmean), and median power frequency (Fmedian). However, these parameters present limitations and principal component analysis (PCA) appears to be an adequate alternative. In this context, we propose two indexes based on PCA to enhance the quantitative muscle fatigue analysis during cyclical contractions. Signals of vastus lateralis muscle were collected during a maximal exercise test. Twenty-four subjects performed the test starting at 12.5 W power output with increments of 12.5 W⋅min–1, maintaining cadence of 50 rpm until voluntary exhaustion. The epochs of myoelectric activation were identified and used to estimate the power spectra. PCA was then applied to the power spectra of each subject. The standard (ST) and Euclidean (ED) distances were employed to estimate the alteration occurred due to fatigue. For comparison, the standard indexes were calculated. ST, ED, and RMS value were adequate for muscle fatigue analysis. Among these parameters, ST was more sensitive with higher effect size. Moreover, the Fmean and Fmedian were not sensitive to fatigue. The proposed method based on PCA of EMG in frequency domain allowed producing fatigue indexes suitable for cyclical contractions.