Abstract
We propose a hand shape recognition system using multiple-viewpoint images. Most conventional hand shape recognition system did not concern about self-occlusion problem due to the hand rotation and a user needs to pay attention to his hand's direction to avoid the problem. We employ multiple-viewpoint images to estimate the pose of a human hand. After a pose estimation, a "best view" for the hand shape recognition is selected based on the estimation result. Hand shape recognition is performed based on the shape representation using P-type Fourier descriptor that is not affected with image scaling and translation. Based on the shape recognition, we developed a system where a user co create virtual graphical scenes interactively. In the system, a user can change the virtual objects' positions, sizes, colors, etc. with hand gestures. This system can be used as a user interface device, replacing glove-type devices and overcoming most of the disadvantages of contact-type devices.
