Visual Perception and Directionality for Modelling and  Evaluating Artificial Crowd and Agent Behaviour

Christopher Peters

Coventry University

Friday, July 15  - 10.30am

Sala Stringa, FBK-irst, Povo

The orienting and directionality of one's body parts,  for example the eyes, head and torso, is a behavior fundamental to the directing, elicitation and signalling of attention in humans. Such behaviours are therefore also worthy of attention themselves as a means for inferring the intentions of others, undoubtedly of great value in survival and social situations. This seminar will discuss relationships between visual perception and directionality from two complimentary perspectives and applied to virtual agents: Firstly, the use of light-weight, real-time synthetic mechanisms to create basic forms of agent perception to actively drive their orienting behaviours in a manner more consistent with internal demands and external stimuli, real and virtual. Secondly, issues concerning human perception of the directionality of humanoid agents (individuals and crowds) for influencing ratings of credibility and task performance, based on the context in which they are perceived. The impact of the attentive loop implied by the aforementioned will be discussed and evaluated with respect to the overall aim of creating virtual agents capable of engaging in more natural and appropriate interactions with humans.

Bio

Christopher Peters is a Senior Lecturer in Computer Games Technology at Coventry University. He obtained his PhD degree in Computer Science from Trinity College Dublin in 2004 and has worked as a post-doctoral researcher at the LINC Communication laboratory, University of Paris VIII and the Graphics Vision and Visualisation Lab, Trinity College Dublin. He is a founding member of the Interactive Worlds Applied Research Group (iWARG) and a research fellow at the Serious Games Institute (SGI) in Coventry University. His research interests include the investigation of interrelationships between computational visual attention, affect, memory, theory of mind and gaze control models for real-time animation, perception of, and multimodal interaction with humanoid characters, groups and crowds.