Researchers from Cornell and the University of Wisconsin–Madison have designed a wrist-mounted device and developed software that allows continuous tracking of the entire human hand in three dimensions.
The research team views the bracelet, called FingerTrak, as a potential breakthrough in wearable sensing technology with applications in areas such as mobile health, human-robot interaction, sign language translation, and virtual reality.
The device senses and translates into three-dimensional coordinates the many positions of the human hand using three or four miniature, low-resolution thermal cameras that read contours of the wrist.
“This was a major discovery by our team — that by looking at your wrist contours, the technology could reconstruct in 3D, with keen accuracy, where your fingers are,” said Cheng Zhang, assistant professor of information science and director of Cornell’s new SciFi Lab, where FingerTrak was developed. “It’s the first system to reconstruct your full hand posture based on the contours of the wrist.”
Yin Li, assistant professor of biostatistics and medical informatics at the UW School of Medicine and Public Health, contributed to the software behind FingerTrak.
“Our team had developed a computer vision algorithm using deep learning, which enables the reconstruction of 3D hand from multiple thermal images,” Li said.
A research article describing the work “FingerTrak: Continuous 3D Hand Pose Tracking by Deep Learning Hand Silhouettes Captured by Miniature Thermal Cameras on Wrist,” was published in June in the Proceedings of the Association for Computing Machinery on Interactive, Mobile, Wearable and Ubiquitous Technologies. It also will be presented at the 2020 ACM International Joint Conference on Pervasive and Ubiquitous Computing, taking place virtually Sept. 12-16.
Past wrist-mounted cameras have been considered too bulky and obtrusive for everyday use, and most could reconstruct only a few discrete hand gestures. Conventional devices have used cameras to capture finger positions.
The FingerTrak device is a lightweight bracelet, allowing for free movement. It uses a combination of thermal imaging and machine learning to virtually reconstruct the hand. Four miniature, thermal cameras – each about the size of a pea – snap multiple silhouette images to form an outline of the hand.
A deep neural network then stitches these silhouette images together and reconstructs the virtual hand in 3D. Through this method, researchers are able to capture the entire hand pose, even when the hand is holding an object.
Zhang said the most promising application is in sign language translation.
“Current sign language translation technology requires the user to either wear a glove or have a camera in the environment, both of which are cumbersome,” he said. “This could really push the current technology into new areas.”
Li suggests that the device could also be of use for health care applications, specifically in monitoring disorders that affect fine-motor skills.
“How we move our hands and fingers often tells about our health condition,” Li said. “A device like this might be used to better understand how the elderly use their hands in daily life, helping to detect early signs of diseases like Parkinson’s and Alzheimer’s.”
In addition to Zhang and Li, the FingerTrak team includes three collaborators who were visiting undergraduate students to Cornell’s SciFi Lab last fall: first author Fang Hu of Shanghai Jiao Tong University; Peng He of Hangzhou Dianzi University; and Songlin Xu of the University of Science and Technology of China.
