Wavelet-Based Signal Processing for Monitoring Discomfort and Fatigue

Abstract

Abstract : Muscle fatigue involves both a decrease in the frequency and increase in the amplitude of a surface electromyographic (SEMG) signal. Muscle fatigue is also related to a decrease of the force impeding to reach the same initial level of the maximum voluntary contraction (MVC). To determine indices of muscle fatigue, a method is proposed to estimate both the instantaneous frequency (IF) and the instantaneous amplitude (IA) by decomposing the SEMG signal using a filter bank. A linear regression model was adopted to compute the IF and IA slopes. These slopes were then classified in muscle increase force, recovery, muscle decrease force and fatigue by using a joint analysis of frequency and amplitude. SEMG signals were recorded from 26 normal human subjects when doing an exertion of 70% and 100% of their MVC during a session of eight hours. It was found that slopes derived from the proposed filter bank are equivalent to those slopes derived from the spectrogram and the smoothed pseudo Wigner-Ville distribution. Furthermore, slopes derived from the filter bank indicated that they can be used as indices to determine muscle fatigue. These results were confirmed by correlating indices of muscle fatigue with perceived levels of discomfort reported by the subjects after performing an exertion of 70% MVC in hours two, four, and six.