Video-based Virtual Hand Input
(H. Sun)

The use of hand gestures is a common means of nonverbal interaction in reality, and can also provide an attractive alternative to cumbersome interface devices. In virtual environments, bare-hand interaction can be much less intrusive than techniques requiring the user to wear special gloves or other uncomfortable devices. Gestures applied to HCI can be helpful to accomplish tasks including pointing & drawing, navigation in virtual environments, manipulation of virtual objects, control of home electronic devices (e.g. TV, CD-player) and robot, and communication through signs. This project is to develop a system that can track & recognize continuous movement of hand/fingertips through real-time video images. Three phases can be identified in the process of gesture analysis, prior to recognition. These are hand localization/modeling, feature extraction, and parameter computation. It is necessary to observe temporal evolution (i.e. whole sequences of frames acquired by the video camera) for dynamic gestures. The system to be developed supports a new input method for wireless & mobile devices.