SEMANTIC CONTENT ANALYSIS OF BROADCASTED SPORTS VIDEOS WITH INTERMODAL COLLABORATION
Naoko Nitta · 2003
This thesis investigates the problem of efficiently describing sports videos for effective multimedia applications, and is summarized as follows: Chapter 1 states the motivation and background of this research and its related works. Chapter 2 discusses the problem of what to describe, addressing both syntactic and semantic structures of the sports videos. Considering the sports videos as a sequence of recurrent semantic story units, we propose a semantic description model for a sports video to represent itself based on its story. The latter part of this thesis discusses how to describe the videos. The automatic generation of the descriptions requires 1)temporal segmentation of videos, and 2)semantic understanding of each segment. Since the several streams included in the video stream can be used collaboratively, this thesis proposes a method of using both the closed-caption text, which is the transcription of the broadcasted video speech, and the image stream, by compensating for limitations in both. Chapter 3 proposes a method for automatically making a description of a play and its related players for each story unit as elements of the proposed semantic description model.