Federated Learning-Enhanced Sensor Fusion for IoT-Based Distributed Health Systems: Prioritizing Privacy and Rapid Medical Intervention

Abstract

Sensor-based Internet of Things (IoT) architectures continue to revolutionize distributed health monitoring systems because of their advancements. The systems enable continuous data collection using wearable and ambient sensors, which promotes precise diagnostic abilities and rapid intervention opportunities. The expansion of distributed health data produces difficulties in privacy preservation and duration of computational processes, and combination accuracy levels during the deployment of traditional centralized machine learning approaches. The privacy preserving federated learning (FL) approach emerged to solve these problems by conducting distributed model training without needing raw data exchanges. The integration of heterogeneous signals depends heavily on sensor fusion techniques to create valuable diagnostic features. The proposed work presents a sensor fusion system based on federated learning designed specifically for healthcare systems operated through the Internet of Things infrastructure. The system architecture allows intelligent diagnostics capabilities and user privacy protections through the combination of heart rate, along with oxygen saturation and body temperature, and electrocardiogram data inputs. Model convergence benefits from a system that uses signal reliability-based adaptive weighting and implements dynamic local update schedules to speed up the fusion process. The approach implements a distributed network structure consisting of independent edge computing devices executing feature identification alongside global model improvement functions. Numerical modeling demonstrates that the implemented system delivers strong diagnostic performance with constrained bandwidth usage and fast model optimization. The diagnostic latency improves substantially when we compare the proposed model with both centralized systems and non-federated techniques. This research offers a comprehensive end-to-end architectural solution alongside mathematical sensor fusion techniques for federated learning, as well as experimental data validation through simulated health information and precise performance measurement reporting.