A Robust and Fast Monocular-Vision-Based Hand Tracking Method for Virtual Touch Screen

An articulated hand tracking method which can be used in virtual touch screen system (VTSS) using monocular camera is proposed. The VTSS is a novel vision-based interactive surface without the need for physical touch screen. The articulated hand model provides more information than fingertips to be in touch with the interactive surface. Using a simplified kinematic model of the hand, a rough estimation for automatic initialization is obtained using the fingertips´ orientation and position under hypothesis of open hand. Then pose of the palm is refined and updated using line features by Gauss-Newton method and re-computation of fingers´ pose based on particle filter followed. Experiments show the potential of the proposed method in applications embedded in VTSS.