Design and Evaluation of a Low-Cost Edge IoT System for Elderly Fall Detection

Abstract

This paper presents the design and controlled experimental evaluation of a low-cost, edge-based elderly health and safety monitoring system implemented on an ESP32 microcontroller. The system integrates a DS18B20 temperature sensor and an MPU6050 inertial measurement unit (IMU) to perform real-time temperature monitoring and threshold-based fall detection. Unlike cloud-dependent approaches, the proposed architecture performs on-device processing to enable low-latency emergency alerting while simultaneously transmitting data to the ThingSpeak cloud platform for remote monitoring. Experimental validation was conducted under controlled simulated fall and activities-of-daily-living (ADL) scenarios. Temperature measurements demonstrated a mean absolute error (MAE) of 0.05°C compared to a clinical reference thermometer. Fall detection performance achieved 85.56% sensitivity, 90.67% specificity, and an overall accuracy of 88.48%. The computed precision and F1-score were 91.67% and 88.48%, respectively. The average fall-to-alert response latency was 0.20 seconds, confirming the effectiveness of real-time edge processing. While the system is validated under controlled conditions, results demonstrate the feasibility of low-cost embedded architectures for rapid emergency detection in home-based monitoring scenarios. The study provides practical performance benchmarking and highlights design trade-offs in threshold-based fall detection on resource-constrained IoT nodes.