Video acquisition on live hypermedia

Live hypermedia is a new type of hypermedia which can acquire a desired set of video scenes from television broadcasting programs using the video scene recognition techniques. Information nodes and links between them care dynamically determined corresponding to the structure of the recognition models. In this paper, a state-transition type expression is proposed to describe flexible and compound models for recognition on live hypermedia. The system can detect not only TV program categories but also their contents. The state transition type model is evaluated in these aspects: retrieval efficiency, code reusability, and computing time during recognition. These experimental results reveal flexibility and effectiveness of our approach.