Video program clustering indexing based on face recognition hybrid model of hidden markov model and support vector machine

Authors:
Yuehua Wan;Shiming Ji;Yi Xie;Xian Zhang;Peijun Xie
Affiliations:
Zhejiang University of Technology, Hangzhou, P.R.China;Zhejiang University of Technology, Hangzhou, P.R.China;Hangzhou University of Commerce, Hangzhou, P.R.China;Zhejiang University of Technology, Hangzhou, P.R.China;Zhejiang University of Technology, Hangzhou, P.R.China
Venue:
IWCIA'04 Proceedings of the 10th international conference on Combinatorial Image Analysis
Year:
2004

Citing 5
Cited 1

A training algorithm for optimal margin classifiers

COLT '92 Proceedings of the fifth annual workshop on Computational learning theory
A multilingual, multimodal digital video library system

Proceedings of the 2nd ACM/IEEE-CS joint conference on Digital libraries
A wearable digital library of personal conversations

Proceedings of the 2nd ACM/IEEE-CS joint conference on Digital libraries
Content-Based Video Indexing for the Support of Digital Library Search

ICDE '02 Proceedings of the 18th International Conference on Data Engineering
Content-based audio classification and retrieval by support vector machines

IEEE Transactions on Neural Networks

Semantic event detection in structured video using hybrid HMM/SVM

CIVR'05 Proceedings of the 4th international conference on Image and Video Retrieval

Quantified Score

Hi-index	0.00

Visualization

Abstract

Human face is a very important semantic cue in video program. Therefore, this paper presents to implement video program content indexing based on Gaussian clustering after face recognition through Support Vector Machine (SVM) and Hidden Markov Model (HMM) hybrid model. The task consists of following steps: first, SVM and HMM hybrid model is used to recognize human face by Independent basis feature of face apparatus; then, the recognized faces are clustered for video content indexing by Mixture Gaussian. From the experiments, the precision of the mixed model for face recognition is 97.8 percent, and the recall is 95.2, which is higher than the complexion model. And the precision of the face clustering indexing is 94.6 percent of the mixed model for compere new program. The indexing result of clustering is famous.