Intelligent delivery of personalised video programmes from a video database
Résumé
The authors present a system for intelligent access to videos, which enables one to apply content-based and cooperative queries to the video database, and to retrieve sets of sequences and personalized videos. The system is based on the previous segmentation of videos into semantic entities. Filters enable one to collect semantic entities into views, which can be collections or scenarios. They can be based both on low-level queries (on image or sound content, on annotations), on complex queries with temporal predicates and on context-dependent and mood-dependent profiles. The system adapts the delivered videos to the user's behaviour. Several implementations are presently being developed within the French Transdoc project.