Personalized Federated Learning for Wearable Time-Series Health Monitoring: Methods, Challenges, and Future Directions
Authors: Zhou, J., Tian, Q., Wang, S., Wang, Y., Yue, X., Yu, H.
Journal: Proceedings of Isoirs 2026 Moving Towards Embodied Intelligence in the AI Age 2026 6th International Symposium on Intelligent Robotics and Systems
Publication Date: 01/01/2026
DOI: 10.1109/ISOIRS70157.2026.11545308
Abstract:Personalized Federated Learning (PFL) enables user-adaptive modeling without centralizing sensitive data, making it suitable for wearable sensing applications. Wearable time-series data are inherently Non-IID due to inter-user variability, device-environment heterogeneity, and evolving physiological states, which limits the effectiveness of a single global federated model. This survey reviews recent advances in PFL for wearable time-series and organizes existing methods into a four-dimensional taxonomy: Non-IID handling, resource and communication efficiency, privacy and security, and label-efficient learning. We also summarize representative datasets and application domains, highlight key design trade-offs, and discuss open challenges toward adaptive, lightweight, and trustworthy personalization for long-term health monitoring.
Source: Scopus