EchoMotion: Towards Full-Body Pose Tracking with a Single Smartwatch Using Active Acoustic Sensing

Full-body pose tracking on wearables enables fitness tracking, rehabilitation, and activity logging on the go, but remains challenging as it requires capturing movements across the torso, arms, legs, and head. Existing solutions often rely on specialized suits or multiple sensors, limiting practicality. We present EchoMotion, an active acoustic sensing system that enables full-body pose tracking using only a wrist-worn smartwatch. By leveraging the built-in speaker and microphone to emit inaudible ultrasonic signals and process their reflections, EchoMotion employs a lightweight deep learning pipeline to infer the 3D positions of 12 body joints. We first validated the approach with 10 participants using a custom hardware prototype, achieving a mean per-joint position error (MPJPE) of 5.36 cm and rotation error (MPJRE) of 11.68°. We then confirmed feasibility with 15 participants on a commercial smartwatch (MPJPE 5.99 cm, MPJRE 11.92°). Finally, in a semi-in-the-wild study with 10 participants, EchoMotion maintained comparable performance (MPJPE 6.41 cm, MPJRE 13.18°), suggesting its potential for future real-world use.