SEMANTIC FEATURES BASED NEWS STORIES SEGMENTATION FOR NEWS RETRIEVAL

In order to find desired video clips efficiently, the research on content-based video retrieval techniques has become one of the most prominent research areas. A multiple semantic features based news stories segmentation approach is proposed in this paper. A prototype system with the capability of the news stories segmentation, and browsing & retrieval is developed for testing the proposed approach. In this approach the video features, (i.e. anchor-person face) and the audio features (i.e. the silence gap and change of speaker) in the news video are detected and used to segment the news stories along with text information (i.e. extracted caption from the news video). The experimental results demonstrate that the proposed approach has higher segmentation precision than that of the caption-based method.