Semantic content representation model for sports video and automatic story segmentation

摘要

Focusing attention on the structure of the broadcasted sports video, we present a model to represent the video in a semantical way by structurizing it and giving the suitable semantical descriptions to its particular time locations. Considering the speech transcript as the useful information stream to help the automatic generation of the proposed semantical descriptions, we try to segment the speech transcript into the units of the structurization in a probabilistic framework based on Bayesian networks. Moreover, the association with the image stream enables us to find the corresponding video segments. In this paper, we discuss some experimental results.