Continuous Physical Activity Assessment through Contactless Posture Sensing

Led by Xinyu Zhang at UC San Diego, this project develops RFPose, a system that uses Wi-Fi devices to reconstruct 3D posture without subject- or environment-specific training. By combining physics-based ray tracking with advanced generative models, the team generates training data to drive deep learning–based wireless posture sensing. The goal is continuous, contactless assessment of fine physical activities to detect deviations in patterns, support caregiver alerts, and reduce data collection costs. A pilot study will recruit participants from UCSD facilities and potentially the AITC network.