An Emotional Model for Nonverbal Communication Based on Fuzzy Dynamic Bayesian Network

Abstract

Autonomous virtual humans play an important role in dynamic virtual environment. Recent research in psychology and neuroscience has identified both the critical role of emotion in decision-making and social interaction. Due to this, to simulate virtual human's believable social communication, a personality-emotion model should be constructed based on artificial psychology first. There are two crucial problems to creating agents that display believable emotions which are how to construct model for emotion, and how emotion model to be applied in agent for improve its performance. In this paper, fuzzy dynamic Bayesian network which is a modification of the dynamic Bayesian network (DBN) is adopted to solve the two questions. Fuzzy logic system enhances our capabilities of capturing the intrinsic uncertainty of human common-sense reasoning and decision-making during their communications. Moreover, dynamic Bayesian network is used to construct the human's personality-emotion model to simulate the transformation of human's emotion. Nonverbal communication is adopted to present our virtual human's emotion change. Nonverbal communication is crucial to increase the believability of virtual humans, during which virtual humans respond not only to the presence of others but also to their postures. An application of this model to create virtual humans involved in social interactions in a virtual community is presented. The interlocutors express their emotion change to one another through their gestures and distance zones. By some experiments of virtual human's nonverbal communication, this system is proved to be able to increase the believability of virtual human to a great extent