Realtime Hand Tracking

by Mick

Researchers from the MIT Computer Science and Artifical Intelligence Lab have designed a system that can recognise gestures made with a multi-coloured gloved hand.

You can see the accuracy and latency of the tracking in the following video


One feature of the system is an algorithm which was developed to rapidly search graphics data in a database.

A webcam is used to capture an image of the glove. The image is reduced in size to 40 x 40 pixels and used to search a database containing digital models of a gloved hand in a range of different positions. If a match is found, the corresponding hand position is determined directly, thus eliminating the need to perform complex calculations to determine the relative positions of the hand and fingers.

Calibration, for different users, has been simplified so that the user only has to place their hand on an 8.5×11 inch piece of paper on a flat surface in front of the webcam.

More details at

https://people.csail.mit.edu/rywang/hand

and from MIT news at

https://web.mit.edu/newsoffice/2010/gesture-computing-0520.html