Kivorent News

Wednesday, November 18, 2026

Health techNews

Cornell team tracks full-body movement using only a smartwatch's speaker and mic

Cornell researchers built EchoMotion, which bounces inaudible sound off the body from a stock Galaxy Watch to estimate the position of 12 joints, no camera needed.

Pulse, covering wearables and health sensors
, 3 min read

2.4 inaverage joint error after a 12-minute calibration

What to know

  • EchoMotion uses a smartwatch's built-in speaker and microphone as a sonar to estimate full-body pose.
  • On an unmodified Galaxy Watch 6, average error per joint was under four inches for new users, about 2.4 inches after calibration.
  • The same model ran on a Galaxy Watch 7 and Pixel Watch 3 without retraining, according to Cornell.

Researchers at Cornell University have shown that an ordinary smartwatch can estimate the pose of a person's whole body, from head to ankles, using only the speaker and microphone already inside it. The system, called EchoMotion, sends out ultrasonic sound that the wearer cannot hear and reads the echoes that bounce back off the body, a form of sonar that the team calls active acoustic sensing. A lightweight AI model turns those echoes into the positions of 12 joints.

The work comes from Cornell's SciFi Lab, directed by Cheng Zhang, an associate professor of information science. It was presented at the ACM UbiComp/ISWC conference on wearable and ubiquitous computing in Shanghai on Oct. 13, and Cornell described the results on Nov. 11. The figures below come from Cornell's announcement; we could not open the full paper.

How it was tested

The team first trained the model with a 3D-printed watch replica fitted with two speaker and microphone pairs, one on each side. Ten volunteers wore it on their non-dominant hand while doing 18 exercises, and the predicted joint positions were compared with positions measured by cameras.

Next came an unmodified Samsung Galaxy Watch 6 and 15 new volunteers. For people the model had never seen, the average error per joint was under four inches. A one-time, 12-minute calibration cut that error by more than a third, to about 2.4 inches, according to Cornell. The same model also worked on a Galaxy Watch 7 and a Google Pixel Watch 3 without retraining.

A third test moved out of the lab. Ten more participants did the 18 exercises in a small room, a narrow hallway and next to a noisy road. Furniture and background noise did not throw the system off, Cornell said, though accuracy dipped slightly in the hallway, where echoes from other people got in the way.

Power draw

Battery life is usually what kills always-on sensing on a watch. Cornell says the sensing drew about 12 milliwatts, roughly 40 times less than a wearable camera, and a 30-minute session used 5% to 6% of a Galaxy Watch 6's battery. "This is a very important jump for body tracking," said lead author Saif Mahmud, a doctoral candidate.

What it could be used for

The team is now pointing EchoMotion at fitness tracking: counting reps and checking form, with the Bulgarian split squat given as an example. It also plans to pair the watch with first-person cameras on smart glasses to follow objects a person is holding. Cornell lists workouts, daily activity and physical rehabilitation tracking as possible uses once the method is built into wearables or virtual reality systems. It is a research prototype, not a feature in any shipping watch, and no company has announced plans to use it. The National Science Foundation funded the work.

Today's watches already infer a lot from their motion sensors, from step counts to fall detection, and they read the heart through light and electrodes, as in our explainer on how smartwatch ECG works. But the accelerometer only knows what the wrist is doing. Seeing the rest of the body has needed cameras or extra sensors strapped to the legs and torso.

Why it matters

If acoustic sensing holds up outside a research setting, the watches people already own, such as the Galaxy Watch 8 line and Pixel Watch, could judge exercise form or track rehabilitation movements with a software update rather than new hardware or a camera pointed at the user.

This article is general information about how technology works. It is not medical advice. Talk to a clinician about your own health.

Sources

WearablesDigital health and AI in careGoogle

Written by an AI correspondent under standards set by editor Jacob Knight. Spotted an error? Tell us and we will fix it.