期刊名称：International Journal of Advanced Research in Computer Engineering & Technology (IJARCET)
印刷版ISSN：2278-1323
出版年度：2013
卷号：2
期号：4
页码：1347-1351
出版社：Shri Pannalal Research Institute of Technolgy
摘要：Nowadays, there are many kind of video such as educational movies, multimedia movies, action movies and scientific movies etc. The careful analysis and classification of movies are very important. For example, movies containing violent or profanity by putting a separate class as they are not suitable for children can cut and avoid from movies for watching. This system is proposed to provide indexing and retrieving the most interesting and important events of the action movie to attract the viewer by using the classified audio categories. The propose system is extracted audio features and make the model by using audio feature vector and classify the audio class to detect and recognize video scenes. In this proposed system, SVM is combined with HMM based on audio features which often directly reflects while image information may not yield useful "action" indexes, to be in detection scene by labeling happy scene, miserable scene and action scene etc. Sound event types are classified including gunshot, scream, car-breaking, people talking, laughter, fighting, shouting and crowd background.
关键词：Audio Indexing; Feature Extraction; ; Hidden Markov Model (HMM); Support Vector Machine ; (SVM).
