Summary
This forum post introduces an arXiv paper (arXiv:2608.20322) by Anton Lambrecht, Reda El Hail, Xianjun Jiao, and Pieter Crombez that provides a controlled comparison of ceiling-mounted FMCW radar, IR-UWB, and Wi-Fi sensing for non-contact health monitoring. The authors collected synchronized recordings from 20 participants across six room layouts and evaluated all three radio technologies using the same CNN on two tasks: fine-grained 10-class human activity recognition and coarse-grained 4-class sleep monitoring. IR-UWB achieved the best cross-subject activity recognition performance (89.0% macro F1), while FMCW generalized best to unseen room layouts (83.8% macro F1). For sleep monitoring, all technologies exceeded 92% macro F1 in unseen environments. The paper identifies a fundamental trade-off between recognition performance and environmental robustness, explained by differences in range resolution, antenna diversity, Doppler resolution, and spatial information preservation. The study addresses an underexplored area: ceiling-mounted deployments, which are attractive for practical installations and cost, are rarely compared under identical conditions.
Paper Overview
Research Area: Machine Learning
Authors: Anton Lambrecht, Reda El Hail, Xianjun Jiao, Pieter Crombez
Published: 2026-08-22
arXiv: 2608.20322
Summary
Although RF-based contactless health monitoring is increasingly important, different radio technologies (FMCW radar, IR-UWB, and Wi-Fi sensing) are rarely compared under the same deployment conditions. Moreover, the performance of ceiling-mounted radar — which offers advantages for real-world installations and cost — remains underexplored.
This paper presents a controlled comparison of FMCW, IR-UWB, and Wi-Fi sensing using synchronized recordings of 20 participants across six room layouts. All technologies were evaluated with the same CNN on:
- Fine-grained 10-class human activity recognition
- Coarse-grained 4-class sleep monitoring
Key Findings
- IR-UWB performed best for cross-subject activity recognition (89.0% macro F1)
- FMCW generalized best to unseen room layouts (83.8% macro F1)
- For sleep monitoring, all technologies exceeded 92% macro F1 in unseen environments
- Results reveal a fundamental trade-off between recognition performance and environmental robustness, explained by differences in range resolution, antenna diversity, Doppler resolution, and spatial information preservation
---
*Auto-collected on 2026-08-22*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178633791