They create a bracelet capable of tracking the posture of the whole body in 3D, and using only one camera

We could see this technological advance in our future smart watches, and that’s how it works.

It is quite common to see millions of people have a watch intelligent or an activity bracelet on their wrists, devices that have advanced a lot in recent times, but are still unable to track the posture of our entire body, although they could end up doing so in the future.

Now Cornell University researchers have created a wristband capable of tracking full-body posture in 3D, solely using a miniature camera and also a custom deep neural network.

Is named Body Trak and it is an intelligent detection device capable of monitoring the user’s body mechanics in all their physical activities.

“Since smartwatches already have a camera, technology like BodyTrak could understand the wearer’s pose and provide feedback in real time. It is practical, affordable and does not limit the area of ​​movement of the user”, points Cheng Zhang, assistant professor of data science and lead author.

BodyTrak, similar to a bracelet, consists of a coin-sized camera on the wrist, along with a personalized deep neural network behind it. This neural network reads the images from the camera or silhouette of the user’s body in motion, and is capable of virtually recreating 14 body poses in 3D and in real time.

Its camera points towards the user’s body collecting only partial images

To challenge and test the wristconducted a user study with nine participants where each of them performed 12 daily activities such as walking, sitting or doing exercise in different settings, wearing different clothes, and with a different number of camera setups.

Then the results after the exercise showed that the system can infer the full body pose, specifically 14-joint 3D positions, with an average error of 6.9 cm using only a miniature RGB camera on the wrist pointing towards the body.

For those who care about privacy, they clarify that their camera points towards the user’s body, collecting only partial images, something that we could see in future bracelets of exercise.

Previously, this research group had already developed deep learning models to track hand and finger movements, but also facial expressions and silent speech recognition.