Multi-Modal Elderly Monitoring via Wearable Devices and WiFi-Enabled Edge-Cloud Architecture with HRV-Based Fall Detection

Abstract

In elderly health monitoring scenarios, events such as falls and cardiac arrhythmias are often sudden and high-risk, making traditional single-indicator monitoring methods insufficient for real-time and accurate detection in complex environments. To enhance system stability and anomaly detection capability under multi-interference conditions, this study develops a wearable monitoring system based on WiFi communication technology. The system integrates physiological signal acquisition, environmental parameter sensing, and dynamic threshold and weight adjustment mechanisms, achieving low-latency and high-reliability data processing and alert response within an edge-cloud collaborative architecture. Experiments were conducted using the SisFall Dataset and the MIT-BIH Physiological Signal Database. The K-Nearest Neighbors (KNN) classifier is employed for anomaly detection. The results showed that under optimal parameter configuration, the system achieved a maximum F1-score of 0.94, peak sensitivity of 94.8%, minimum event detection latency of 140 ms, and a minimum false alarm rate of only 3.8%. The simulation tests further verify the robustness and adaptability of the system under low, medium, and high interference environments, demonstrating its feasibility for deployment in smart elderly care and remote healthcare applications.