Neuromuscular Regulation Mechanisms and Recovery Strategies under Exercise-Induced Fatigue: An Empirical Study

Abstract

Exercise-induced fatigue alters muscle activation, intermuscular coordination, and movement stability, making reliable fatigue assessment important for training-load regulation and recovery planning. However, conventional approaches often rely on isolated surface electromyography (sEMG) indicators or classification models that provide limited representation of intermuscular regulation and weak linkage between fatigue recognition and recovery-oriented decisions. This study proposes a Neuromuscular Fatigue Regulation Network (NFR-Net) integrating temporal–spectral sEMG representation, movement information, adaptive intermuscular coupling, confidence calibration, and graded recovery-priority assessment. Experiments were conducted using publicly available fatigue-related sEMG datasets with subject-level data separation and five independent runs to ensure robust evaluation. NFR-Net achieved a balanced accuracy of 90.3 ± 1.8% and a Macro-F1 of 0.897 ± 0.020, while its AUROC of 0.946 ± 0.013 was slightly lower than the 0.949 ± 0.016 obtained by BiLSTM. External validation of the intermuscular coupling mechanism showed moderate agreement between coupling-based muscle rankings and participant-reported fatigue (ρ = 0.58), with stronger agreement with physiotherapist assessment (ρ = 0.64). Furthermore, confidence calibration significantly reduced the expected calibration error (ECE) from 0.076 ± 0.014 to 0.039 ± 0.009, enhancing model reliability. These findings indicate that combining local electrophysiological features with muscle-level coordination can support interpretable fatigue monitoring and more reliable recovery-priority decisions, ultimately facilitating optimized athletic training and rehabilitation protocols.