Researchers at Cornell have created EchoMotion, a system that leverages ultrasonic sound from a smartwatch's speaker and mic to reconstruct full-body pose with under‑4‑inch joint error, requiring no additional hardware.
How EchoMotion works on smartwatches
EchoMotion operates like sonar: the watch speaker emits ultrasonic pulses while the built‑in microphone records the echoes bouncing off the wearer’s body. All processing is performed in software, eliminating the need for external sensors.
The AI model powering the system is under 30 MB and draws roughly 12 mW, allowing it to run on Wear OS devices without noticeable impact on battery life or system resources.
Accuracy and testing results
In the initial study, ten participants performed 18 different movements while wearing a prototype watch. The model predicted the positions of 12 major joints—from ankles to head—with an average error just under 4 inches.
Cross‑device performance
The same model was deployed on an unmodified Samsung Galaxy Watch 6, then tested with 15 new users, yielding the same sub‑4‑inch error. After a 12‑minute personal calibration, the error dropped to about 2.4 inches. The system also ran on Galaxy Watch 7 and Google Pixel Watch 3 with only a slight accuracy dip, indicating compatibility with older generations.
Limitations and future applications
Accuracy fell when participants moved in a crowded hallway, suggesting interference from nearby people. Furniture clutter and street noise had little effect, highlighting the primary challenge of multi‑user environments.
Potential for XR and rehabilitation
The team is exploring integration with smart glasses to enable full‑body gesture input for XR experiences. In healthcare, the detailed joint data could support remote monitoring of rehabilitation exercises, offering timely feedback to patients.
Power consumption and practical demo
During a 30‑minute session on the Galaxy Watch 6, EchoMotion consumed only 5‑6 % of the battery, confirming its low‑power design. A demo app recognized 20 gym workouts—including squats and dumbbell lateral raises—with 95.4 % accuracy, even in gyms where background music reached 76 dB.
TechTimes Perspective
TechTimes believes EchoMotion could reshape how Vietnamese consumers track workouts, as it removes the need for costly add‑on hardware. If manufacturers integrate the software into upcoming models such as the Galaxy Watch 9 or Pixel Watch 5, users will gain full‑body pose tracking for the price of a standard smartwatch. The main hurdle remains sensitivity to nearby people, which may limit accuracy in crowded gyms initially, but the low power draw and hardware‑free approach make it a promising candidate for mass adoption.