A research group at Cornell University has developed a special bracelet capable of tracking the posture of the entire body in 3D, thanks to a miniature camera and a neural network.
BodyTrak, this is the name of the device, is an intelligent sensing technology that could make a huge difference in monitoring the body mechanics of physical activity: a field in which precision is fundamental.
According to Cornell researchers, BodyTrak is the first bracelet capable of doing this using a single, small camera.
A bracelet will scan us
“Technology like this, which “understands” the user's posture and gives them real-time feedback, is very important,” he says Cheng Zhang, senior author of the study. “By using a single camera, this bracelet does not restrict the user's movements.”
Further on, I believe such a detection system will be integrated in the next generations of SmartWatch.
How does it work?
BodyTrak's neural network collects camera data to read the body silhouette and recreate a double of it. Indeed, 14: there are many 14 3D poses that the system can process in real time.
To test the bracelet, the researchers conducted a study involving 9 participants, that I link to you here. Each of the participants performed 12 daily activities (walking, sitting, exercising, etc.) in different scenarios, with different clothes and different settings.
The results? Crazy. BodyTrak infers whole body pose (3D positions of 14 joints) with an average error of 6,9 cm. And it does this using a single miniature RGB camera (11,5mm x 9,5mm) on the wrist pointing towards the body.
What about privacy?
“Our research shows that body frames do not need to be completely within the camera's view to be detected,” said Hyunchul Lim, co-author of the study. “By capturing even just a small part of our body, this bracelet can “duplicate” it entirely in 3D.”
This may mean that such a device can also be used to scan the body of anyone in its vicinity. And this raises legitimate concerns.
The researchers respond that this bracelet does not present privacy issues, since the camera is pointed towards the user's body.
IF it's pointed at the user's body, I say.
In conclusion, such a system could have a thousand applications: our wrist devices will tell us when we walk badly, or if we are recovering quickly from an accident. But caution will still be needed in the development of a technology like this.