Remote Health Monitoring via Federated Learning on IoT Devices for Elderly Care

This study proposes a new federated learning (FL)-based remote health monitoring framework specifically designed for elderly care using Internet-of Things (IoT)-enabled wearable devices. The designed system provides a real-time solution for privacy-aware monitoring of physiological signals that capture heart rate, blood pressure, and mobility patterns without having to shift any raw data to a centralized server. Instead of transferring the raw data, each device will train a local model while sending only the updates of the model, which will later be aggregated using secure approach to offer a global predictive model. As the global model is distilled and executed every few minutes and retrained afterwards with another global model updated, it therefore ensures both personalized and adaptive remote health monitoring for medicine or health care in forming physiological states as just illustrated with sleep patterns. Experimental results confirm the proposed method outperforms existing centralized and edge approaches in terms of accuracy (93.7%) and F1-score (92.4%) and with a significant communication overhead, while improving privacy. In contrast to the state-of-the-art approaches, the proposed framework is proven more scalable, efficient, and secure for continuous health monitoring, such that it enables early anomaly detection, and timely preventive action among aged populations, if necessary. Finally, the proposed approach allows for easier wide-scale deployment in a real world healthcare setting