Towards real-time speech-based facial animation applications built on HUGE architecture

This paper examines how our existing HUGE architecture aids fulfilling requirements emerging from the everyday expansion of new multimedia applications. Trends in humancomputer interaction stream towards creating natural user interfaces which apply rules from human-human interaction. Virtual humans who look natural, and behave believably fit perfectly in the concept of creating natural user interfaces. Universal architecture for statistically based HUman GEsturing (HUGE) is a system for producing and using statistical models for facial gestures based on any kind of inducement. We present the general architecture and its adoption to speech-based facial gesturing, and further justify it through the analysis of requirements for building multimedia application. We believe that this universal architecture is useful for experimenting with various kinds of potential applications since it enables automatic generation of full facial animation in the real time.