Northwestern studies teach smartwatches to read hand gestures and new activities
Northwestern's SPICE Lab showed two smartwatch systems at UIST 2026: EITWatch reads gestures from muscles and tendons, and TransfHAR learns a new activity from one minute of motion data.
Pulse, covering wearables and health sensors
, 2 min read
92%EITWatch accuracy on six common hand gestures
What to know
- EITWatch uses eight electrodes on the back of a watch to sense muscles and tendons, and recognized six common gestures with over 92% accuracy, according to Northwestern.
- TransfHAR, built on more than 2,700 hours of public motion data, learned seven new activities at about 90% accuracy from a one-minute recording of each in a 10-person study.
- Both are research prototypes; Northwestern says the gesture work could one day help people control prosthetic hands.
Researchers at Northwestern University have shown two ways a smartwatch could understand what a wearer's hands and body are doing. One system, EITWatch, reads the muscles and tendons in the wrist to recognize hand gestures. The other, TransfHAR, uses the motion sensors already in a watch and can learn a new activity from about a minute of examples.
Both come from Northwestern's SPICE Lab, led by Karan Ahuja, an assistant professor of computer science at the McCormick School of Engineering. The studies were presented at the ACM Symposium on User Interface Software and Technology (UIST) 2026, and Northwestern described them on Nov. 4. The figures below come from Northwestern's announcement; we could not open the papers themselves.
EITWatch: gestures from the wrist
EITWatch uses a technique called electrical impedance tomography (EIT). Eight electrodes on the underside of the watch pass tiny currents through the wrist and measure how they change as muscles and tendons move, building a rough picture of what the hand is doing.
In testing, the prototype recognized six common gestures, such as a thumbs up, a clenched fist and pointing, with more than 92% accuracy, according to Northwestern. It recognized five subtler finger movements with more than 91% accuracy. It works best when calibrated to one person, and Northwestern calls it "still a research prototype." Xuanyou Liu and Novel Alam of the SPICE Lab contributed equally to the work.
Northwestern says this kind of sensing could eventually let someone steer a prosthetic hand, or control devices with a gesture, from a watch rather than a dedicated armband.
TransfHAR: new activities from one minute of data
Activity tracking on today's watches covers a fixed menu of workouts that the maker trained ahead of time. TransfHAR tries to let the watch pick up new ones on the spot. It reads the watch's built-in accelerometer and gyroscope, the same sensors that power step counting and fall detection, and was pretrained on more than 2,700 hours of public motion data.
In a study with 10 participants, the system learned seven activities it had not seen before at about 90% accuracy after a one-minute recording of each, according to Northwestern. The co-authors include SPICE Lab PhD student Aidan Bradshaw, Riku Arakawa of Carnegie Mellon University and Xin Liu of Google Health Research. Northwestern says Google partners are helping scale up the data.
Where this fits
University labs have been pushing the sensors in stock watches well past what they were built for. This month Cornell described a watch estimating full-body pose using only its speaker and microphone, which we covered in Cornell's EchoMotion study. Neither Northwestern system is in a product, and no company has announced plans to ship them. EITWatch also needs electrodes that current watches lack, while TransfHAR runs on sensors every smartwatch already has.
Why it matters
If methods like TransfHAR reach shipping software, a watch such as the Galaxy Watch 8 or Pixel Watch could track a physical therapy routine or an unusual sport after a minute of setup, and wrist-based gesture sensing could give people with limb differences a lighter way to control a prosthetic hand.
This article is general information about how technology works. It is not medical advice. Talk to a clinician about your own health.
Sources
- Smartwatch tech from Northwestern's SPICE Lab could help control prosthetic hands, Northwestern Now, Nov 4, 2026
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.