A Shirt Might Replace Your Fitness Tracker

It may seem out of character for a techie to track their workouts with a pen and paper. Between smartwatches and the many other fitness trackers available today, there are plenty of options for keeping tabs on your exercise habits without lugging around a notebook. However, these devices tend to be worn on the wrist, and that limits how accurate the data they collect is. The wrist is a fine location for monitoring one’s heart rate or step count — but it is going to let you down in a big way on leg day at the gym.

To capture finer details, more sensors need to be distributed around the body. But given the option of looking like they are wearing a motion capture suit or are a Luddite with a notebook, most people are going to choose to go the low-tech route. A new system developed by researchers at Cornell University makes it possible to capture body-wide motion data without any visible sensors, so we may no longer have to choose form over function. Using it is as easy as putting on your shirt.

Called SeamFit, the team’s new smart clothing system looks, wears, and even washes like an ordinary t-shirt — but it is embedded with conductive threads that can detect posture and exercise movements in real-time.

To make sense of the data produced by a SeamFit shirt, a machine learning pipeline interprets changes in the threads’ capacitance — their ability to store electrical charge — as the wearer moves. These changes are picked up by a small, detachable circuit board at the back of the shirt’s neckline and transmitted via Bluetooth to a computer for processing. From there, SeamFit can identify specific exercises and count repetitions — all without any manual input from the wearer.

In a user study involving 15 participants performing 14 common exercises such as lunges, sit-ups, and bicep curls, SeamFit successfully classified the exercises with 93.4% accuracy and counted repetitions with an average error of less than one rep. Notably, it required no calibration or individual training, and it continued to perform after being washed and dried.

This is not the team’s first foray into smart textiles. SeamFit builds on their earlier project, SeamPose, which used a long-sleeve shirt to track posture. But where SeamPose required more seam sensors, SeamFit refined the approach, using just a few carefully placed seams to achieve similar results in a more practical way.

The practical applications of SeamFit stretch beyond just the gym. The team envisions potential future applications in physical therapy, remote health monitoring, and even human-AI interaction. Because it can accurately recognize a user’s activity, SeamFit could help AI systems better understand when and how to engage with people in everyday life.

Currently, the researchers are working on ways to scale up production using industrial serger sewing machines and more durable conductive threads. They believe this will lead to a future where smart clothing that is indistinguishable from regular wear will be a part of our everyday lives.