Content-Based Audio Video Searching And Commercial Detection

With the fast development of Internet and the rapid expansion of data volume, Search technology has become the first way for people to get informations. People are not limited to only use words as the entrance of search. The development of content-based audio-video searching makes using audio or video as searching entrance possible, and this filed has become a very promising technology.In this paper, A new content-based audio searching framework is proposed. The main work is as follows:1.In the audio feature extraction stage, the audio frames were filterd by a large amount of targeted filter, which greatly eliminated the distortions caused by seven kinds of audio conversion, and proposed to improved audio feature named FWASF.2.In the building audio searching index stage, use Locality Sensitive Hash technology and inverse index methods to build a high-efficiency index structure.3.In the audio matching stage, proposed a weighted temporal-pyramid matching algorithm, and then use the new level algorithm fusioned separate result of the audio frame to thereby obtain a final audio searching result.Experiments on TRECVID datasets show that the results generated by this proposed framework has achieved a leading level of the world.With the development of the Internet, advertising distribution channel has spread from the traditional TV channel to the Internet platform. A peculiarity of Internet advertising is the fast speed of transmission. How to collect, regulate these advertisement effectively has become an urgent problem. This paper proposed an commercial automatic recognition and commercial segmentation system based on learning method. The main work and innovation are as follows:1.Proposed a text detection method based on the Harris point and decision tree, which can detect the text area in advertising video effectively. Then design a variety of features to describe the text area, which is a powerful supplement of the original advertisement detection features.2. Basd on the shot segmentation, proposed a feature extraction method based on the adjacent shot, and a variety of post-processing methods are proposed to improve the detection results.3. Analysed the difference of the advertisement and non-advertisement from the audio aspect, and has obtained certain conclusions.