利用情感激励提取足球视频精彩镜头-数字媒体处理与检索实验室.pdfVIP

ISSNN 111777TP Journal of Computer Research and Development 7( 10) : 2010 1, 2 1 1 于俊清 何欢欢 何云峰 1 ( 3007 ) 2 ( 3007 ) ( yjqing@ hust. edu. cn) Highlights Extraction for Soccer Video Based on Affection Arousal 1, 2 1 1 Yu Junqing , He H uanhuan , and H e Yunfeng 1 ( School of Comp uter S cience Technology, H uaz hong University of Science and Technology , Wuhan 3007 ) 2 ( Center of N etw or and Comp utation, H uaz hong Univ ers ity of Science and Technolog y, Wuhan 3007 ) Abstract In recent years, the football video retrieval and abstract techniques develop more and more rapidly. As one of the key technologies of video retrieval and abstract for soccer video, highlights extracting technique has high academic research value and w ide application perspective. Based on the analysis of existing football video research methods, it is proposed that a highlights extracting system using affection arousal modeling from the angle of audiences emotional change, which is based on the theory of Hanjalics affection curve. T he arousal features are extracted from football video to construct arousal modeling and generate arousal curve, then highlights can be extracted based on arousal curve. Compared w ith Hanjalics w ork, some improvements and further research are proposed based on Hanjalics excitement curve. A new feature, named shot intensity, is designed to construct arousal modeling instead of the motion intensity feature. T his has improved the systems recall rate, accuracy and computing performance. Domain know ledge is applied in the highlights shot locating process to achieve high accuracy, and the highlights can also be automatically filtered a