Optimisation and comparison framework for monocular camera-based face tracking

Tracking the position and orientation of the human face with respect to a camera has valuable applications in human computer interaction (HCI). Examples are navigating through a virtual environment, controlling objects using head gestures, and enabling avatars in a virtual environment to reflect the user's behaviour. Tracking performance can be heavily influenced by environmental parameters. Developers and users of face tracking plugins without computer vision experience need guidelines how to optimise face tracking performance in real world set-ups and they need measures how environmental parameters influence the results. In this paper we develop a qualitative framework for determining ideal working conditions of face tracking algorithms. We apply our framework to a commercially available face tracking solution and present the results of this analysis.