Digital content manipulating system using hand gesture

In this paper, a real-time stereo visual system which is capable of recognizing user's gesture is presented. In the proposed system, the image stream of user is captured by two off-the-shelf webcams to calculate the depth information which is used to detect the user's position and head. The operating area of user's hand is dynamically determined based on the head position of user. A circular template is applied to search the center point of user's palm, which would be corrected by the directional weighting. The trajectories formed by the motion of user's hand are smoothed by averaging and B-spline curve fitting. Finally, several basic gestures are tested by 12 participants to verify the performance of the proposed system.