A Portable Scene-Aware View Management System for Video-Conferencing

We present a scene recognition framework for a novel automatic view management system in video conferencing. The system aims to provide the best viewpoints to the viewer based on the integration of both audio and video information. Distinguished from previous approaches, our system is aware of the ´scenes´ going on, thus applying an appropriate view management policy for different scenarios is made possible. In this paper, we describe the framework of the view management system, with emphasis on the scene recognition based view management and the portability of the system. Preliminary experiments are performed for the evaluation of our approach.