Estimation of muscle fatigue degree using time-varying autoregressive model parameter estimation of surface electromyography

Abstract

Objective: Aiming at investigating the nonstationary characters of the EMG signal, time-varying AR model was employed in this study to quickly estimate muscle fatigue by analyzing short time surface electromyography. Methods: Data from 10 subjects pre- and post-fatigue was analyzed by the AR model. A recursive least squares algorithm was then used to extract the time-varying parameters and transformed the time-varying question into time-stable one. Results: The first time-varying parameter shows higher sensitivity to fatigue than that of the traditional median frequency (sensitivities increased from 37.80% to 324.46%). Conclusions: The mean value of the first time-varying parameter could be used as a fast indicator to reflect the fatigue of the muscle, which would promote practical applications in the field of lumbar muscle fatigue diagnosis and rehabilitation. Also it would provide a reliable tool for the study of ergonomics.