Abstract. This paper addresses the problem of estimating head pose over a wide range of angles from low-resolution images. Faces are detected using chrominance-based features. Grey-level normalized face imagettes serve as input for linear auto-associative memory. One memory is computed for each pose using a Widrow-Hoff learning rule. Head pose is classified with a winner-takes-all process. We compare results from our method with abilities of human subjects to estimate head pose from the same data set. Our method achieves similar results in estimating orientation in tilt (head nodding) angle, and higher precision for estimating orientation in the pan (side-to-side) angle.

Abstract

Abstract. This paper presents a novel approach to detect interaction groups in intelligent environments. To understand human activity, we must identify human actors as well as their interpersonal links. Interaction detection is a good cue to address activity of user groups. An attentional model is derived from gravitational model and cognitive psychology approaches. Whereas determining locally users'.focus of attention is a difficult task, this model exploits contextual elements such as position, speed and saKency of objects in the scene to estimate shared attention. The attentional model shows promising results on simulated scenarios where unexpected events occur.

RESUMEDans cette démonstration, nous présentons une nouvelle génération de kiosque interactif intégré, multipoints et réagissant à la présence de personnes, destiné à l'usage dans un espace public. Les systèmes interactifs publics existants n'offrent qu'une interaction très spécifique, ce qui limite leur utilité, leur utilisabilité et leur attractivité. Pour résoudre ces défauts, nous proposons une solution de tactilisation fluide et de perception du contexte fondée sur la vision par ordinateur. Nous couplons cette perception à une boîte à outils de construction de systèmes interactifs, inspirée des architectures à agents PAC. Une application interactive est décrite de manière similaire à une application web (ressources XML, asynchrone, CSS, et un langage similaire à JavaScript). Le rendu graphique est effectué par le canevas GML. Des installations commerciales montrent que les utilisateurs naïfs s'approprient et utilisent notre kiosque. ABSTRACTIn this demo, we present a new integrated, multi-touch and presence-aware kiosk, designed for public spaces. Existing public interactive systems offer limited interactions, which hinders their utility, usability and attractivity. To overcome these obstacles, we propose a computer vision-based approach to fluid tactilisation and context perception. This is coupled with an interaction toolkit inspired from PAC agents achitectures. An interactive application is described much like a web application-using asynchronous XML, CSS, and a Javascriptlike language. GUI rendering performance is achieved by using the OpenGL-based canvas GML. Commercial installations show that naive users appropriate and use our kiosk.