Research On Video Semantic Content Analysis

With the rapid development of computer networks and multimedia technology, digital video plays an important role in our daily life. However in those videos, there exists some bad information, which seriously and harmfully affects the society. Therefore, video content security gets the great place in our social life and national security.Video semantic analysis is an important research work in digital video content security. As a high-level semantic clue of video, the content of audio in digital videos provides valuable information for people to understand the semantic content of the video. Algorithms for audio classification, audio segmentation and speech recognition is meaning for video semantic analysis.This thesis presents an audio classification and speech recognition framework for video semantic analysis. Firstly, an audio classifier based on rule and Support Vector Machines (SVM) is designed. Three audio classes are considered: silence, music and speech. Four rules are presented and applied in the final audio segmentation. Then an automatic speech system based on CMU Sphinx toolkit was built. Experiments on video from TRECVID 2005, the results show that the efficiency of the framework.