Head tracking for interactive computer graphics has been explored for over 30 years in general, and 20 years at UNC. Having finally achieved sufficient performance in the laboratory, researchers have begun to explore methods for tracking with reduced infrastructure. A common and yet elusive goal is to reduce or virtually eliminate the user-worn component and the environmental infrastructure, while simultaneously maintaining the high performance of current laboratory-based systems. One exciting application is Augmented Reality beyond the laboratory: virtual images and information visually attached to physical objects in the real world all around us.
After a brief review of previous head tracking research, I will describe our high-performance HiBall Tracking System, attempting to convey some intuition about why it works so well. I will discuss what I believe is the most significant problem facing high-performance pose estimation and sensing in general: understanding and managing system-wide sensor information and uncertainty. I will then describe some preliminary ideas for a mathematical and graphical intelligence-amplification framework aimed at achieving sensor information efficiency for pose estimation. Finally I will speculate on how this work might extend to other applications including tele-immersion and new ideas for projector-based graphics.