A method for extraction of audio-visual leitmotif in movies by cross media analysis

The joint analysis of video and audio components in multimedia documents has been widely used since the beginning of activities related to the new multimedia Standard MPEG7. In this context the paper is focused on a method for extraction of an emotional and structuring cue from artistic content we call “audio-visual leitmotif”, which is a first step in characterization of an author style in producing the content. The method is based on joint motion -based video partitioning and on model-based music recognition.