Take the Kinect and drop it in the hands of the geeky, but exceptionally talented folks at MIT and the result is nothing short of impressive – Minority Report impressive. As reported by Kotaku, a group at MIT coded up some hand detection software capable of distinguishing your hand and fingers within a cloud of 60,000 points and at a rate of 30 FPS per seconds. Add in a simple but sweet graphical user interface, the open source robotics package ‘ROS, and the libfreenect driver that lets the Kinect interface with Linux and you have the groundwork for a real-life Minority Report interface.
The group at MIT put together a two-minute video (shown below) that showcases this novel interface. With this system, a test user is able to scan through an image gallery by moving their hand in a circular motion. The user is able to grab an image and bring it to the forefront where he can zoom or rotate it at will. When done with the image, the user merely pushes it off the screen. The system is able to pinpoint the fingers, which appear as five white dots on the screen and the palm which appears as a pink blob. Of course, real-life performance with this Kinect-based system is not as fluid as the movie, but it is an excellent start.
The work was spearheaded by the Learning and Intelligent Systems (LIS) group and the Robot Locomotion group, both of which work out of the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT. They have provided the code behind this project, tutorials, and a mailing list if you are interested in learning more about how this system works.