Abstract
There is a complex unwritten code which regulates human interactions. In this paper we present a camera based monitoring system that explores the relationship between proxemics, visual attention and personality traits during interaction. People's relative positions an head poses are extracted with a multi-target tracking algorithm and used to (i) estimate, under the'thin slices' hypothesis, the level of extroversion and neuroticism of a person, to (ii) learn a model of people interactive behavior that improves, once integrated in the algorithm, the accuracy of the tracking estimates, and to (iii) steer a set of active cameras to the subject found to exhibit the most peculiar interaction pattern. We report on experimental results in a natural scenario where people engage in a party.