Video-based interaction for a mixed reality kiosk system

Mixed Reality Kiosk Systems used for information access in public places facilitate easy and intuitive Human Computer Interaction even for technically unversed users. Therefore, natural conversation with anthropomorphic avatars as the most intuitive form of interaction particularly requires speech and gesture input and output. In this paper we focus on aspects of video-based input modalities like region tracking, gesture recognition and pattern recognition that are implemented in a prototype of a physical information booth.