Real-time stereo matching for 3D hand gesture recognition

3D hand gesture recognition is an emerging technology as the market demands new user interfaces of electronic devices, especially for gaming, mobile devices, and home entertainment applications. We proposed a stereo matching design which is optimized for the 3D hand gesture recognition. As most of the existing 3D gesture recognition systems require customized camera devices (e.g. IR camera and IR light source of TOF technology or fine-tuned stereo-camera) to deliver quality results, our real-time stereo matching system can be configured to support various stereo-cameras from different vendors and produce stable depth map. Besides, we also demonstrate the real-time 3D hand gesture recognition with our stereo matching design in this paper. The experiment results show that our stereo matching design is efficient to process full-HD video stream such as 1080i side-by-side video standard and generate real-time depth map at 60 fps. Our 3D hand gesture recognition not only allows users to play virtual mouse on the PC but also recognizes delicate 3D motion such as the finger clicking on the virtual bottom.