Web-based Chinese News Video Semantic Content Security Analysis

Today, due to the variety of the multimedia information content on the Web, and its information security attracts more and more attention. Video semantic content security analysis is an important and hot topic. Compared with generic video data, news video has the greatest influence on people daily life. Now news video processing has been widely investigated, and the content semantic extraction of the video is the first step for video processing and video understanding. Supported by the National Natural Science Foundation of China, this thesis studies on the techniques of content semantic analysis of Web-based Chinese news video.As for the lower accuracy of recognition of video ASR, this thesis proposes an improved framework for the semantic analysis of the Chinese news video. Firstly, we segment the news video into series of story units employing a two-level approach of news story segmentation. Secondly, we extract the key phrases from the corresponding ASR transcript of news story using an improved TFIDF algorithm. Thirdly, we calculate the Pinyin similarity to find the semantic keywords,using the similarity of spelling of the video ASR transcript . Finally, the more precise content semantic of the video from Web via the search engine was obtained from the relevant web pages on the Web by employing the method that filters the topical content.The experimental results on several hours Chinese news video (CCTV4_NEWS) from the TRECVID2005 show that our framework is feasible .The framework will be helpful for news video semantic content security, automatic abstraction and access control with XML documents as well in the future work.