A gesture recognition system using stereo vision and arm model fitting

When one uses a traditional motion capture system for gesture recognition, colored or infrared markers need to be attached to the moving body. It takes extra time and effort and increases the cost of the devices. If markers are not used, it is very difficult to distinguish the arm area from a variable background. In order to achieve markerless gesture recognition that avoids such a difficulty, we propose a new gesture recognition algorithm and system that uses stereo vision and arm model fitting. Because our algorithm is very robust with a variety of clothing worn by the recognized person and also a varied background, it can be used in daily life and in various applications.