Construction of a Muscle Fatigue Estimation Model Considering Physiological Background Using Bayesian Network

ABSTRACT

This paper attempted to relate muscle fatigue, muscle action potential (EMG: Electromyogram), metabolic activity generated in the body, and vital signs such as respiration and heart rate by using a Bayesian network (BN), considering that they are caused by interrelated stochastic transitions. The BN was constructed based on EMG, heart rate, respiratory rate, and subjective muscle fatigue obtained from four male subjects in their 20s who performed three sets of arm curl‐like exercises for 4, 5, and 6 kg loads, respectively. As a result of providing each node of the constructed BN with known evidence, the posterior probability in the node representing subjective muscle fatigue changed as expected. Thus, the results suggest that the BN can represent the relationship between biologically relevant signals and muscle fatigue, including uncertainty. Modeling with this method has the potential to estimate muscle fatigue without requiring expertise in EMG interpretation.