One of the most important projects taken by CAMS research is the development of algorithms and software for motion capture and interpretation. This direction targets applications in intelligent, human-computer interface for computer and video games, automation of animation, computer based athletic coaching (golf, tennis), remote surgery, and much more. For brevity, this technology will be referred to as VRS (Virtual Reality System).
VRS can be considered as a set of computer vision algorithms for interactive graphics with applications and potential benefits for computer games, surgeons, disabled patients, sports enthusiasts, educators, and numerous others. In particular, users can, through hand gestures, manipulate objects (a ball and a butterfly, for example, have been used in prototypes), and they can control machines or appliances. The interactive applications pose the following challenges and constraints:
The algorithms should be real-time: no appreciable delay should exist between when a user makes a gesture or motion and when the computer responds.
The algorithms should be reliable: they should work for different subjects and against various backgrounds.
Interfaces and software should be affordable and cost effective (economic constraint).
CAMS developed and applied the computer vision algorithm suitable for interactive graphics applications. The algorithm consists of the following three fundamental sub-systems: shape recognition, motion analysis, and object tracking. The innovative VRS software allows for the interpretation of human movements as control commands, which are understood by a computer. VRS provides a computer with the ability to capture, recognize, and interpret as commands the gestures, motions, and postures of human operators.
The main elements of VRS are as follows:
A fast color segmentation and object registration algorithm which enables the computer to track the landmarks in real time.
A proprietary gesture recognition algorithm which enables the computer to understand human gestures and act accordingly.
An intelligent game interface which feeds the visual gesture input to various games.
CAMS investigators include Boris Rozovsky and George Yaralov.
Movies
Here you can download a 30 second 320x240 movie demonstrating the VRS technology.
Demo
To see our technology in action, please download a self-extracting demo (1 MB).
System requirements:
Pentium 166MHz MMX or higher.
16 MB memory.
Video-for-Windows compatible video camera capable of delivering 320x240 frames at 10 frames per second (QuickCam and parallel port based cameras are a little bit slow; cameras with dedicated video capture boards are much faster).
Please let us know what you think about our technology at yaralov@scf.usc.edu
