A group of scientists from Cornell University (USA) proposed an intelligent system BodyTrakthat could appear in smartwatches in a few years – this system uses a coin-sized RGB camera and an artificial intelligence-based algorithm to track the position of its owner’s entire body.
Worn on the wrist, the camera transmits the image to a neural network that is trained to reconstruct the position of the entire body from fragmented images. The system works in real time and creates a 3D model of the human body based on 14 positions.
BodyTrack was demonstrated with nine volunteers who wore the camera during simple daily activities such as walking, sitting or exercising. A camera pointed at the owner’s body was enough to recreate three-dimensional models with an average error of 6.9 cm.
In practice, such a system can replace the expensive devices currently used for motion detection in the professional field. It can also be used in smartwatches to monitor sports activities or the user’s general condition in real time. However, the developers found that modern smartwatches do not yet have the necessary cameras or computing components powerful enough to run the neural network.