Implementing Gaze-corrected Videoconferencing

Lack of gaze awareness is a key failure that hinders the widespread acceptance of videoconferencing. GazeMaster is a project which attempts to provide a software solution to gaze awareness and eye contact. Previous publications have described the general approach of GazeMaster. This paper describes our implementation experience, giving more details of our software architecture, and explaining our approach to head modeling. Our results with respect to video, graphics, and networking are solid. Our computer vision technology, while promising, is still immature, and requires further research.

[1]  David G. Novick,et al.  Coordinating turn-taking with gaze , 1996, Proceeding of Fourth International Conference on Spoken Language Processing. ICSLP '96.

[2]  Kentaro Toyama,et al.  A Warp-Based Feature Tracker , 1999 .

[3]  Gregory D. Hager,et al.  Incremental Focus of Attention for Robust Vision-Based Tracking , 1999, International Journal of Computer Vision.

[4]  Kentaro Toyama,et al.  Gaze-awareness for Videoconferencing: A Software Approach , 2000 .

[5]  Rogério Schmidt Feris,et al.  Hierarchical wavelet networks for facial feature localization , 2002, Proceedings of Fifth IEEE International Conference on Automatic Face Gesture Recognition.