Unsupervised TV program boundaries detection based on audiovisual features

We evaluate in this paper how the hypothesis that basic video and audio features present homogeneous values during programs can be exploited by a GLR-BIC segmentation algorithm to identify programs in days of television contents. The homogeneity criterion is evaluated by the ability to describe this feature values with a Gaussian law. On the base of results obtained from rough audio and video features, this paper evaluates the improvement obtained by an early fusion of these features, and by a prior usage of the existence of monochromatic frames in commercials.