A segmentation method of news video stories based on announcer’s voiceprint

As an important step of content based news video retrieving and intelligence mining, semantic unit segmentation has attracted many researcherspsila interests. This paper focuses on a new method of news video stories segmentation which is based on the announcerspsila voiceprints. Firstly, the voiceprints included acoustic perception characteristics of all announcers have been extracted, and its Gaussian mixture model will be trained, then the audio clips included announcers and not announcers will be detected by the KL divergence method, at last the semantic units will be segmented under the guidance of video topic caption frames events and special structure knowledge of news program. Finally the 92.58% recall and the 96.02% precision have been achieved during more than 2 hourspsila experiment.