Abstract
The purpose of this research is to develop an interface which enables us to operate TV's by hand motions. We use a stereo camera to recognize a hand motion. A fast image processing is necessary to recognize quick hand motions. So we always track a person in front of a TV by using upper body in depth images and face, and process only an area surrounding him/her. The fingertip position is detected as the most protruded region in the depth image around the faces. In gesture recognition, the trajectory of the fingertip are analyzed to classify the hand motion into a predetermined set of gestures. In pointing recognition, we calculate the corresponding between the position of the fingertip and that on the display. The relationship between the coordinate system in the image and that on the display enables to mapping the fingertip position onto the display.
