Learning to recognize affective body postures

Abstract

A --Robots ore assuming an increasingly important role in our society. They now become pels and h e b support children healing. In other words, they are now hying lo enterlain an active and affective communieotion with human agents. However, up to now, such systems hove primarily relied on the hwnan ogsnts'nbilfly to empathize wifh the system. Chnngcr in tho behavior ofthe system could therefofore resull in changes of mood or behavior in the humon p a r " But current system do not seem 10 read lo users, or only in clearly pre-defined ways. In thal Sense, curr~nt system miss the bi-directionality typical to human social interaction. Social interaction is characterized by (1 multi-channel communication, in which each actor captures and reacts to signals by the other actor. To this aim, a compaler or a robol has to be oble to capture and interpret sign& senf by the human partner in order to achieve socinl interaction. One ofthe most important channels ofcommunication is physical internctim. lhle body h used to interpret the nffeective stde ofan interlocutor. This paper describes erperiments we carried (1-1 to study the importance of body language in affective communication. The results ofthe erperimenls led us to develop n system that con inerrmenlolly leom to recognize affective stotcs from bodypostures.