BodyTrak wrist camera constructs 3D models of the body in real time
Wearable technology is capable of tracking various measures of human health and is getting better all the time. New research shows how this could come to mean real-time feedback on posture and body mechanics. A research team at Cornell University has demonstrated this functionality in a novel camera system for the wrist, which it hopes to work into smartwatches of the future.
The system is dubbed BodyTrak and comes from the same lab behind a face-tracking wearable we looked at earlier in the year that is able to recreate facial expressions on a digital avatar through sonar. This time around, the group made use of a tiny dime-sized RGB camera and a customized AI to construct models of the entire body.
The camera is worn on the wrist and relays basic images of body parts in motion to a deep neural network, which had been trained to turn these snippets into virtual recreations of the body. This works in real time and fills in the blanks left by the camera’s images to construct 3D models of the body in 14 different poses.
The researchers demonstrated the potential of its BodyTrack system with the help of nine participants, who wore the camera as they carried out everyday activities like walking, sitting or exercise. With the single camera worn on the wrist and pointed at the body, the system was able to construct full models with an average positioning error of 6.9 cm (2.7 in).
“Our research shows that we don’t need our body frames to be fully within camera view for body sensing,” said lead author of the research Hyunchul Lim. “If we are able to capture just a part of our bodies, that is a lot of information to infer to reconstruct the full body.”
This kind of technology could have potential in next-generation motion capture systems, which traditionally involve expensive equipment to track the complex mechanics of moving bodies. The team imagines it also finding use in smartwatches, where it could track body position during physical activity in real time. They note that this is still a ways off, as the current crop of smartwatches don’t have the necessary cameras and hardware to carry out full body sensing. They are, nonetheless, excited about the potential.
“Since smartwatches already have a camera, technology like BodyTrak could understand the user’s pose and give real-time feedback,” said Cheng Zhang, senior author of the paper. “That’s handy, affordable and does not limit the user’s moving area.”
The research was published in the journal Proceedings of the Association for Computing Machinery.
Source: Cornell University