Sami Huttunen
Machine Vision Prize 2006 Winner

One potential application area of camera-based user interface technology is videoconferencing. A special case of videoconferencing is distance education, where the teacher needs constantly switch the sources of the video feed between the computer, document camera, blackboard camera, overview camera, and possibly other devices. For the teacher, this is annoying, and in practice these actions may easily be forgotten which might deteriorate the effectiveness of teaching and learning. It is obvious that the usability of the system could be improved significantly by changing the sources automatically based on the teacher’s gestures and behaviour.

When developing an automated system for video source switching, there is at least one basic requirement. The system should be reliable in its decisions, because erroneous decisions or constantly changing camera views are irritating from the audience’s point of view. In principle, the decisions should not differ substantially from the decisions made by a human operator. So, the problem for constructing the automated system is twofold. The system needs reliable information as a basis for detecting and understanding the events in the scene, and it must have the knowledge for making the appropriate decisions. In order to have the necessary information, the system needs to extract such features from the video that could describe the possible events requiring some actions. However, the procedure for extracting the features should be computationally inexpensive to guarantee real-time performance.

The system that was developed in the master’s thesis takes care of both the selection and switching of the video source in a distance education situation. The camera views from the classroom are used to observe the teacher’s movements and possible document camera usage. In this case the stimulus for switching the source is obtained directly from the video cameras, which offers an unobtrusive way to keep watch on the person’s actions. Again it is not necessary to have any additional sensors apart from the cameras already existing in a typical distance education classroom system. Rule-based video source selection is made on the basis of both the teacher’s location and document camera activity. The final video source switching is handled by the Videra’s auditorium control unit (ACU) which controls the equipment in the distance education classroom.

The results obtained indicate that the system implemented can clearly provide relief for the teacher when using the distance education system. Since the use of videoconferencing technology in distance education increases all the time, there is a growing need for the solution presented in the thesis.