ETRI-Knowledge Sharing Plaform

ENGLISH

성과물

논문 검색
구분 SCI
연도 ~ 키워드

상세정보

학술지 Automatic Media Data Rating Based on Class Probability Output Networks
Cited 9 time in scopus Download 3 time Share share facebook twitter linkedin kakaostory
저자
Harvey Rosas, 길이만, 한승완
발행일
201011
출처
IEEE Transactions on Consumer Electronics, v.56 no.4, pp.2296-2302
ISSN
0098-3063
출판사
IEEE
DOI
https://dx.doi.org/10.1109/TCE.2010.5681103
협약과제
10MS2100, 유해 멀티미디어 콘텐츠 분석/차단 기술개발, 정병호
초록
This paper presents a novel method of classifying media data whether they include X-rated contents or not. In our work, the classification of media data is performed using the class probability output network (CPON) which estimates the conditional class probability. Consequently, the classification of media data can be done using the degree of confidence for the class membership, not just using the discriminant value which is usually used in many classification problems. Furthermore, the accuracy of the estimated conditional class probability can be measured in the suggested CPON and this gives a good guideline for the final decision of classification. To demonstrate the effectiveness of the suggested method, the simulation for automatic media rating of the data sampled from multimedia data streams in the Internet was performed. We showed that the suggested CPON-based method provides the better performance than other classifiers using discriminant functions . © 2006 IEEE.
키워드
Class Probability Output Network, Media Data Rating, Multimedia Data, Support Vector Machine
KSP 제안 키워드
Class membership, Class probability output network, Classification problems, Conditional class probability, Data stream, Degree of confidence, Discriminant functions, Multimedia data, Support VectorMachine(SVM), novel method