Ambient IoT Backscatter Sensing for Fall Detection and Localization in Smart Healthcare

Authors:
Publisher website
Google Scholar
Abstract
Falls remain a major cause of injury and death among older adults, which shows the need for reliable and non-intrusive monitoring solutions in healthcare environments. In this paper, we propose a novel Ambient Internet of Things (IoT) backscatter sensing system that utilizes a dense array of passive tags and a minimal number of reader antennas for cost-effective fall detection and localization. To fully exploit the spatial and temporal characteristics of ambient backscatter sensing data, we design a hierarchical multi-task spatio-temporal graph attention network (HM-STGAT), which jointly models the spatial relationships among tags and antennas as well as the temporal dynamics of human activities. The proposed unified framework simultaneously detects fall events and accurately estimates fall locations. We validate the proposed approach through a real-world experiment to collect a diverse dataset of fall and non-fall scenarios. Experimental results demonstrate that the proposed method achieves state-of-the-art performance in both fall detection accuracy and localization precision, highlighting its potential for practical deployment in healthcare monitoring applications.
Funding Information
  • Engineering and Physical Sciences Research Council (EP/Y000315/2)
  • National Natural Science Foundation of China (62471208)
  • Natural Science Foundation of Guangdong Province (2026B1515020041)

This publication has 33 references indexed in Scilit: