抄録
We have developed a finger-pointing positional input device for TV operation. The device takes images of a viewer from a triple stereo color camera and detects the viewer's fingertip and face from three-dimensional position and color information. The device then maps the fingertip position to the TV display by using the positions of fingertip and face. The fingertip is detected by subtracting the input image from the background image and taking the closest point in the resulting "human area" to the TV display. The face is detected by extracting a skin color model from the neighborhood of the fingertip and searching for a skin-colored region above the fingertip in the human area.
