Audiovisual integration for tennis broadcast structuring.pdf

AUDIOVISUAL INTEGRATION FOR TENNIS BROADCAST STRUCTURING E. Kijak 1,2 , G. Gravier 2 , L. Oisel 1 , P. Gros 2 1 Thomson multimedia RD, France 2 IRISA 1, av. de Belle-Fontaine Campus Universitaire de Beaulieu 35510 Cesson-Sevigne, France 35042 Rennes Cedex, France {ewa.kijak, lionel.oisel}@ {ggravier, pgros}@irisa.fr ABSTRACT This paper focuses on the integration of multimodal features for sport video structure analysis. The method relies on a statistical model which takes into account both the shot con- tent and the interleaving of shots. This stochastic modelling is performed in the global framework of Hidden Markov Models (HMMs) that can be efficiently applied to merge au- dio and visual cues. Our approach is validated in the partic- ular domain of tennis videos. The model integrates prior in- formation about tennis content and editing rules. The basic temporal unit is the video shot. Visual features are used to characterize the type of shot view. Audio features describe the audio events within a video shot. As a result, typical tennis scenes are simultaneously segmented and identified. 1. INTRODUCTION Video content-based analysis is an active research domain that aims at automatically extracting high-level semantic events from video. The extracted semantic information can be used to produce indexes or table of contents that enable efficient search and browsing of the video content. Indexing tasks usually attempt to identify a predetermined set of events within a video. Producing a table of contents implies to perform a temporal segmentation of the video into shots. Such a task has been widely studied and generally relies on the detection of discontinuities into low-level visual fea- tures such as color or motion [1]. The critical step is to au- tomatically group shots into ”scenes”. Scenes are referred in the literature as story units. They are defined as a coher- ent group of shots that is meaningful for the end-user. The main problem relies on the definition of