Detection of Exercise Fatigue Using Neural Network with Grey Relational Analysis from HRV Signal

摘要

This paper proposes an exercise fatigue detection model based on real-time physiological data, combined with time-domain analysis (time-domain analysis), frequency domain analysis (frequency domain analysis), trend analysis (trend wave analysis, DFA), approximate entropy (Approximate entropy, ApEn) and sample entropy (sample entropy, SampEn). As a result, this research also presented a feature selection with the Analytical hierarchy process (GRA), which extracted vital impact factors. Finally, the Back-propagation Neural Network (BPNN) is used to analyze each data feature to identify the individual’s exercise fatigue. According to the empirical results, the exercise fatigue detection model proposed in this study can effectively distinguish the degree of exercise fatigue and improve the accuracy to a satisfactory level.