Identifying aesthetic highlights in movies from clustering of physiological and behavioral signals

Affective computing is an important research area of computer science, with strong ties with humanities in particular. In this work we detail recent research activities towards determining moments of aesthetic importance in movies, on the basis of the reactions of multiple spectators. These reactions correspond to the multimodal reaction profile of a group of people and are computed from their physiological and behavioral signals. The highlight identification system using the reaction profile is evaluated on the basis of annotated aesthetic moments. The proposed architecture shows significant ability to determine moments of aesthetic importance, despite the challenges resulting from its operation in ecological situation, i.e. real-life recordings of the reactions of spectators watching a film in the movie theater.