AI-powered smart clothing logs posture and exercises

Researchers at Cornell have developed a brand new sort of good clothes that may monitor an individual’s posture and train routine however appears to be like, wears—and washes—similar to an everyday shirt.

The brand new expertise, known as SeamFit, makes use of versatile conductive threads sewn into the neck, arm and facet seams of a typical short-sleeved T-shirt. The consumer doesn’t must manually log their exercise, as a result of a synthetic intelligence pipeline detects actions, identifies the train and counts reps. Afterward, the consumer merely removes a circuit board on the again neckline, and tosses the sweaty shirt into the washer.

The group envisions that SeamFit might be helpful for athletes, health fanatics and sufferers engaged in bodily remedy.

Most current body-tracking clothes is tight and restrictive or embedded with chunky sensors, in accordance with Catherine Yu, a doctoral scholar within the area of data science and lead researcher on the challenge.

“We have been interested by how we will make clothes good with out making it cumbersome or unusable,” Yu mentioned, “and to push the practicality, so that individuals can deal with it the way in which they’d often deal with their clothes.”

Alternatively, athletes can select health trackers, like smartwatches or rings, however these are additional gadgets that individuals might not wish to put on whereas exercising, and might’t monitor motion throughout the complete physique.

“Not everyone seems to be prepared to check out a brand new wearable kind issue, however folks may have garments on,” mentioned co-author Cheng Zhang, assistant professor of data science within the Cornell Ann S. Bowers Faculty of Computing and Data Science. “We offer a really neat kind issue that’s all the time on you.”

Their research,”SeamFit: In the direction of Sensible Sensible Clothes for Automated Train Logging,” published in March within the Proceedings of the ACM on Interactive, Cell, Wearable and Ubiquitous Applied sciences, and will likely be introduced on the UbiComp/ISWC 2025 assembly in October in Espoo, Finland.

Most mass-produced clothes has seams, which Yu realized might be exploited to make a cushty, inexpensive piece of good clothes. She constructed three SeamFit shirts—in small, medium and enormous—utilizing a house stitching machine to connect conductive threads on high of the seams. The three sizes allowed customers to decide on a looser or tighter match, however did complicate the method of deciphering every consumer’s actions.

To check the shirts’ efficiency, the group recruited 15 volunteers, who did a sequence of 14 workout routines—together with lunges, sit-ups and biceps curls—whereas sporting SeamFit. With none calibration or coaching for every consumer, SeamFit’s mannequin categorised the workout routines with 93.4% accuracy and efficiently counted reps, with counts that, on common, have been off by lower than one.

SeamFit works as a result of when folks train, the threads’ capacitance—their means to retailer cost—modifications because the threads transfer, deform and work together with the human physique. The circuit board on the again neckline measures the capacitances and transmits them via a Bluetooth connection to a pc. A custom-made, light-weight signal-processing and machine-learning pipeline then deciphers the actions.

After the exercises, Yu washed and dried the shirts at house.

The challenge is a brand new iteration of SeamPose, a earlier effort to trace physique postures utilizing conductive threads in eight seams of a long-sleeve T-shirt.

The group envisions that this kind of unobtrusive good clothes might be particularly helpful for athletes logging their train routines and for bodily therapists monitoring the progress of sufferers at house.

Extra broadly, this kind of expertise may help with human-AI interplay, as a result of by monitoring human actions and actions, AI can higher perceive when to work together and when to attend—corresponding to when somebody is consuming or sleeping.

Enabling AI to know human exercise is the principle focus of Zhang’s Sensible Laptop Interfaces for Future Interactions (SciFi) Lab, which develops new, AI-powered wearable sensing techniques, to allow AI to understand human actions and intentions in on a regular basis settings and supply assist when wanted.

“Whereas this paper demonstrated the strategy for a easy garment, we imagine it might probably simply be tailored to a variety of clothes and will benefit from the advanced seam patterns of superior sportswear,” mentioned co-author François Guimbretière, professor of data science in Cornell Bowers CIS and the multicollege Division of Design Tech.

To create SeamFit, Yu arrange a “little stitching manufacturing unit” within the lab. Nevertheless, she is presently exploring how the manufacturing course of might be affordably scaled up, utilizing industrial serger machines—which sew and make seams utilizing three or 4 threads concurrently—and extra strong conductive threads.

“By simply changing a single thread on this mass manufacturing course of, the entire clothes may simply develop into good and be capable to have this movement monitoring functionality,” Yu mentioned. “I am imagining at some point, you open your closet and there is actually no distinction between good and nonsmart clothes.”