Highlight Ranking for Racquet Sports Video in User Attention Subspaces Based on Relevance Feedback
Lists
Only you see this
Abstract
Inthis paper, wepropose amethod torankthehighlights of broadcast racquet sports videos. Compared withprevious work, weintegrate relevance feedback into highlight rankingframework toeffectively capture theuser's interest in attention subspaces andgenerate personalized ranking result. First, weestablish three userattention subspaces andextract audio, visual, temporal affective features torepresent the humanperception ofhighlight ineachsubspace. Then, the highlight ranking models areconstructed using support vectorregression (SVR)forthethree subspaces respectively. Finally, thethree submodels arelinearly combined togeneratethefinal ranking model. Relevance feedback technique isemployed toadjust theweights ofeachsubmodel toobtain theresult whichissuitable totheuser's preference. Experimental results demonstrate ourapproach iseffective.
BibTeX
@inproceedings{zheng2007highlight,
author = {Yijia Zheng and Guangyu Zhu and Shuqiang Jiang and Qingming Huang and Wen Gao},
title = {{Highlight Ranking for Racquet Sports Video in User Attention Subspaces Based on Relevance Feedback}},
booktitle = {ICME},
year = {2007},
doi = {10.1109/icme.2007.4284597},
}Loading related content...