MIT Engineers Build Smart Carpet To Track Workout, Be Your Fitness Coach

MIT researchers have developed a smart carpet that can recognise and sense human movement without any cameras, and track fitness activities in real-time.

MIT

Also Read: MIT Researchers Build Free App To Reduce Indoor Covid-19 Transmission Risks

The intelligent mat was revealed at this week¡¯s Conference on Computer Vision and Pattern Recognition conference. At the demo, engineers showcased a 36x2 feet mat that was able to register the posture and movement of the individual on it and their relationship to the ground.

The mat has been made using commercial-grade, pressure-sensitive films as well as conductive threads and has over 9,000 sensors that gauge pressure from the feet, limbs and other parts of the human body and turn it into an electrical signal.

Testing revealed that it was capable of recognising a person¡¯s motion at an accuracy of 97 percent while highlighting the individual¡¯s error in pose in less than four inches.

While systems exist that accurately analyse posture using cameras, they¡¯re not regarded as the most ideal due to privacy concerns, while often having clarity issues.

Also Read: MIT Engineers Build Smart Cloth Sensor That Can Track Our Health And Save Lives

In the creation of this smart carpet, the CSAIL team behind this creation used cameras only to train the mat¡¯s initial dataset. After achieving that, the neural network was smart enough to figure out if someone was lying prone, stretching or whatever pose they were up to, based on the tactile information on the mat.

MIT

But the mat is not perfect. For now, it was able to accurately recognise lower body motions and leg placement, but it failed to accurately determine motions from the top half of the body. Moreover, in case a move lacked direct floor contact, like free-floating legs during sit-ups, the prediction of gestures wouldn¡¯t be that accurate.

Also Read: Ramesh Raskar's MIT Team Builds App That Tells If Someone's Infected With COVID-19 Near You

Co-author Yunzhu Li, a PhD student at MIT, feels this could be incorporated into gaming or for people who are into home workouts, ¡°Based solely on tactile information, it can recognize the activity, count the number of reps, and calculate the amount of burned calories.¡±

Lead author, Yiyue Luo, on the other hand, felt that it could help in monitoring the health of the elderly, physical rehab routines or detecting falls.