Tracking-dependent and interactive video projection

Gestures’ expressivity, as perceived by humans, may be related to the amount of attention they attract. In this paper, we present three experiments that quantify behavior saliency by the rarity of selected motion and gestural features in a given context. The first two ones deal with the current quantity of motion of a person’s silhouette compared to a brief history of his quantity of motion values and with the current speed compared to a brief history of the person’s speed. The third one focuses on the motion speed of a person compared to the motion speed of other persons around him. Considering both features (speed and quantity of motion) and contexts (space and time), we compute an attention index providing cues on the behavior novelty. This can be considered as a preliminary step to an expressive gesture analysis based on behavioral. In order to achieve accurate tracking, a fusion between color and IR camera streams is achieved. This fusion let us have a robust tracking system with respect to illumination and partial occlusions issues.

Il n'y a actuellement aucune contribution dans cette catégorie.