Leveraging FMCW Radar for Monitoring On-bed States of Human Presence, Posture, and Motion

In this study, we propose a non-invasive solution for on-bed monitoring of human states using FMCW radar. Our unique approach models the bed area and region of interest (ROI) from the radar perspective, projecting them into azimuth-range and elevation-range maps for precise state detection. The processing pipeline uses the Minimum Variance Distortionless Response (MVDR) for target localization and Fast Fourier Transform (FFT) and Weighted Range-Time-Frequency Transform (WRTFT) for Doppler data extraction. We introduce a new method for on-bed state detection by capitalizing on chest localization within the azimuth-elevation-range-ROI mapping, enhancing detection accuracy. Our solution concurrently monitors on-bed presence, posture, motion, and turning, providing a holistic view of sleep behavior. Extensive experimental evaluation confirms the solution’s effectiveness, efficiency, and robustness, with an F1-score of 100% for presence detection and similar high scores for posture, motion, and turning detection.