Abstract

Motivation: Helping parents to automatically detect the most violent scenes in a movie to prevent their children watching them. Applying multi-modal analysis (visual and audio) for violence detection in movies. Goals: Detection of violent segments in movies at video shot level. Fusing visual and audio features for video content description and semantic analysis. Evaluation of SIFT local features with Bag-of-Words (BoW) approach for visual feature analysis. Time series analysis of audio signals in movie data.