Real-Time Health Monitoring for Athletes in Dynamic Sports Environments Based on Split Federated Learning and Recurrent Neural Networks
IEEE Transactions on Consumer Electronics · 2025-08 · 71 (3) · 8956-8967
DOI: 10.1109/tce.2025.3587200 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Consumer Electronics, 2025-08. Verfasst von Lijun Tao und 1 weiteren Autor:innen. Die Arbeit wurde bisher 12-mal zitiert (Crossref) und nennt 37 Literaturangaben.
Abstract (Originalsprache)
The rapid advancements in the Internet of Things (IoT) and artificial intelligence (AI) have played a crucial role in transforming real-time health monitoring and sports analytics. However, traditional AI-based systems face challenges in data privacy, computational overhead, and real-time inference. Centralized deep learning requires high resources and poses a risk to data security. At the same time, Federated Learning (FL) improves privacy but remains computationally intensive, struggling with sequential data modeling, which makes it less suitable for dynamic sports environments that require efficient real-time health assessments. To address these limitations, this paper proposed a novel Split Federated Learning (SFL) framework integrated with Recurrent Neural Networks (RNN) for real-time athlete health monitoring. The SFL model partitions training between edge devices and cloud servers, ensuring privacy while reducing computational overhead. The RNN component effectively captures temporal dependencies in physiological signals, enhancing real-time anomaly detection and predictive health analysis. First, wearable IoT sensors collect real-time physiological data, including heart rate, temperature, and oxygen saturation. Secondly, data undergoes preprocessing, feature extraction, and encryption for quality and security before partial edge-device processing. Thirdly, intermediate representations are securely transmitted to the cloud, where RNN layers perform sequence modeling and final classification. Finally, a federated server aggregates model updates from multiple devices using secure aggregation techniques. Experimental results validate SFL-RNN’s effectiveness, demonstrating a 12% reduction in computational latency while maintaining high predictive performance: accuracy of 94.1%, precision of 92.7%, recall of 93.2%, F1-score of 92.9%, MCC of 0.871, and AUC of 0.964, respectively. A comparative analysis with centralized AI, traditional federated learning (FL), and standalone split learning highlights SFL-RNN’s superior privacy preservation, computational efficiency, and real-time inference capabilities, making it a robust solution for professional sports, rehabilitation, and medical diagnostics.
Zitieren
APA
Tao, L., & Li, H. (2025). Real-Time Health Monitoring for Athletes in Dynamic Sports Environments Based on Split Federated Learning and Recurrent Neural Networks. IEEE Transactions on Consumer Electronics, 71(3), 8956-8967. https://doi.org/10.1109/tce.2025.3587200
Vancouver
Tao L, Li H. Real-Time Health Monitoring for Athletes in Dynamic Sports Environments Based on Split Federated Learning and Recurrent Neural Networks. IEEE Transactions on Consumer Electronics. 2025;71(3):8956-8967. doi:10.1109/tce.2025.3587200
BibTeX
@article{Tao2025,
author = {Tao, Lijun and Li, Hai},
title = {Real-Time Health Monitoring for Athletes in Dynamic Sports Environments Based on Split Federated Learning and Recurrent Neural Networks},
journal = {IEEE Transactions on Consumer Electronics},
year = {2025},
volume = {71},
number = {3},
pages = {8956--8967},
doi = {10.1109/tce.2025.3587200}
}
EUVIMED