Despite my mom's continuing devotion to her analog pedometer, wearable sensors that track fitness are now a major technology category, one that largely didn't exist a decade ago. But what if the sensors aren't as useful at tracking some kinds of exercise as a simple video camera?
That's the premise of a project out Carnegie Mellon University, where researchers have found that a stationary camera paired with a machine vision algorithm is superior for tracking gym exercises than, say, your Fitbit. The system is called GymCam, and stationed in a room full of people exercising it detects repetitive motions to track workouts.
"In a gym, the repetitive motion almost always is an exercise," said Mayank Goel, assistant professor in CMU's Human-Computer Interaction Institute (HCII) and Institute for Software Research. "If you are moving both your arms, you tend to move them together in time. However, if two people are exercising next to each other and performing the same exercise, they are usually not in sync, and we can tell the difference between them."
If the thought of a camera watching you while you work out in a room full of strangers is less than appealing, the researchers behind GymCam clearly feel the same way. A spokesperson for CMU emphasizes that the system only needs to track pixel-by-pixel changes and that faces can be eliminated from the capture to reduce or eliminate any privacy intrusions.
Among other fascinating achievements, the researchers figured out how to compensate for partial obstruction of people exercising, a not unlikely challenge for a single-camera system in a gym littered with equipment. GymCam can detect exercise as long as its camera can see any body part moving repetitively.
Rushil Khurana and Karan Ahuja, Ph.D. students at CMU leading the research, recently presented their findings at the Orwellian-sounding International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp 2019) in London. Ahuja explained that smartwatches do a reasonable job of tracking many cardio exercises, but the scope of what they track is limited by the part of the body where they're meant to be worn. A smartwatch, for example, might not sense leg presses and might have a hard time differentiating between various concurrent body motions. A camera, by contrast, is relatively cheap and can, using this algorithm, accurately measure exercises of all kinds, so long as they're confined to a camera's field of view.
The underlying technology has all kinds of possible applications. Perhaps undermining some of the privacy protections the CMU team put in place, one application is to identify people by a unique "movement signature."
The system also might have uses beyond physical exercise. Goel said the camera system, combined with smartwatches worn by individuals, might help people with visual disabilities navigate shopping malls, airports and other public spaces. Instead of using the person's face as their identity, the system will use their motion as their signature. It allows people to easily opt-out of being tracked or located.