Introducing human-like hearing perception in intelligent virtual agents

Abstract

Although, at this moment, there are not many researches on perceptual models for Intelligent Virtual Agents, the few that exist are more focused on visual than on hearing perception. This paper describes the work that the Universidad Politécnica de Madrid, in collaboration with the University of Nottingham, is carrying on, with the aim of endowing Intelligent Virtual Agents with a synthetic model of human-like hearing perception. We have based the design and development of this model on Endsley studies about situational awareness [10,11], where perception can be understood as the first level of awareness. This perceptual model also extends and reinterprets the key concepts introduced by a CSCW model of awareness known as the Spatial Model of Interaction [4]. In this paper, we give an overview of those studies related to our work and we describe how we have achieved our purposes: selecting an agent architecture; re-defining and reinterpreting the set of key concepts introduced by the Spatial Model of Interaction; and analysing those factors that make the auditory perceptual model more realistic. We also expose how a mathematical function can describe the agents clarity of perception and how this model of perception has been implemented and evaluated.