Regression algorithm for emotion detection

International audienceWe present here two components of a computational system for emotion detection. PEMs (Personalized Emotion Maps) store links between bodily expressions and emotion values, and are individually calibrated to capture each person's emotion profile. They are an implementation based on aspects of Scherer's theoretical complex system model of emotion~\cite{scherer00, scherer09}. We also present a regression algorithm that determines a person's emotional feeling from sensor measurements of their bodily expressions, using their individual PEMs. The aim of this architecture is to dissociate sensor measurements of bodily expression from the emotion expression interpretation, thus allowing flexibility in the choice of sensors. We test the prototype system using video sequences of facial expressions and demonstrate the real-time capabilities of the system for detecting emotion. We note that, interestingly, the system detects the sort of hysteresis phenomenon in changing emotional state as suggested by Scherer's psychological model