Adaptation d'une théorie psychologique pour la génération d'expressions faciales synthétiques pour des agents d'interface

In this article, we describe how human-agent interactions may be improved by refined believability of facial expressions of zoomorphic or anthropomorphic embodiement forms. The believability of agents' emotional expressions depends on the strength of the link between the expressions themselves (external states) and the emotional model (internal states or processes). Scherer's psychological theory describes how human facial expressions are linked with internal cognitive appraisal processes. In this paper we show how we adapted Scherer's theory of human emotional facial expressions generation from cognitive appraisal to the generation of artificial emotional expressions on two different embodiement platforms: a graphical avatar and a physical robotic head. We discuss the results of the user study we conducted to assess the believability of the two embodied forms.