mV-IMU: mmWave-enabled Virtual Inertia Measurement Unit for High-fidelity Activities of Daily Living Monitoring

Zhi Zhang, Chongxin Zhong, Yuliang Fu, Ying Chen, Jinyuan Jia, Huining Li, Chenhan Xu

Confirmation: I have read and agree with the IEEE BSN 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: inertial measurement unit, mmWave sensing

TL;DR: We present mV-IMU, a mmWave-enabled Virtual Inertial Measurement Unit framework to measure body inertia without attaching wearable sensors

Abstract: Monitoring human motion through inertial metrics is vital for healthcare, rehabilitation, and activity recognition. Traditional approaches rely on wearable inertial measurement units (IMUs), which, despite their accuracy, impose burdens due to their intrusive nature, limiting long-term usability. To mitigate this, recent advances explore device-free alternatives, such as pose-based inertial inference from video or mmWave sensing. However, inertial signals derived from pose tracking are prone to error amplification during differentiation. In this paper, we present mV-IMU, a novel mmWave-enabled Virtual Inertial Measurement Unit framework that bypasses pose estimation altogether to directly reconstruct body accelerations from raw mmWave signals. Our approach leverages a deep inertia reconstruction model trained on kinematics-informed features extracted from mmWave point clouds, integrated with a physics-guided optimization scheme for enhanced accuracy. Extensive evaluations show that mV-IMU achieves inertial measurement fidelity close to wearable IMUs, enabling practical, non-intrusive motion monitoring for smart healthcare and rehabilitation contexts.

Track: 7. Contact-less solutions for human sensing

NominateReviewer: Zhi Zhang ([e-mail])

Submission Number: 100