📝 SummarXiv

Abstract

Egocentric body tracking, also known as inside-out body tracking (IOBT), is an essential technology for applications like gesture control and codec avatar in mixed reality (MR), including augmented reality (AR) and virtual reality (VR). However, it is more challenging than exocentric body tracking due to the limited view angles of camera-based solutions, which provide only sparse and self-occluded input from head-mounted cameras, especially for lower-body parts. To address these challenges, we propose, BodyWave, an IOBT system based on millimeter-wave (mmWave) radar, which can detect non-line-of-sight. It offers low SWAP+C (size, weight, and power consumption), robustness to environmental and user factors, and enhanced privacy over camera-based solutions. Our prototype, modeled after the Meta Quest 3 form factor, places radars just 4cm away from the face, which significantly advances the practicality of radar-based IOBT. We tackle the sparsity issue of mmWave radar by processing the raw signal into high-resolution range profiles to predict fine-grained 3D coordinates of body keypoints. In a user study with 14 participants and around 500,000 frames of collected data, we achieved a mean per-joint position error (MPJPE) of 9.85 cm on unseen users, 4.94 cm with a few minutes of user calibration, and 3.86 cm in a fully-adapted user-dependent setting. This is comparable to state-of-the-art camera-based IOBT systems, introducing a robust and privacy-preserving alternative for MR applications.