Wrist ping-pong: Cornell's sonar smartwatch trick spies on your whole body
Cornell University researchers built EchoMotion, a software-only system that makes a commercial smartwatch emit inaudible ultrasound and read the echoes with a small AI model. It reconstructs 12 body joints in real time without extra hardware. On a Galaxy Watch 6, new users' joint positions were off by an average of 9.3 cm, falling to 6 cm after 12 minutes of fine-tuning. In gym tests it identified 20 exercises with 95.4% accuracy, though heavy sleeves block the signal.
Why it matters: It suggests ordinary smartwatches could track whole-body movement through software alone, with no extra hardware, which could enable fitness tracking for wearers, though accuracy varies by person and heavy sleeves block the signal.
Comments
No comments yet.