COLLOQUIUM Department of Computer Science and Engineering University of South Carolina Semantic Video Classification and Indexing For Medical Education Applications Jianping Fan Department of Computer Science University of North Carolina at Charlotte Date: October 17, 2003 (Friday) Time: 3:30-4:30PM Place: Swearingen 1A03 (Faculty Lounge) Abstract Digital video now plays an important role in medical education, healthcare, and other medical applications. As large-scale collections of medical education videos come into view, there is an urgent need to classify, summarize, and organize medical education videos at the semantic level. In spite of recent research progress, semantic video classification is still an open problem with many unsolved challenging issues. Under the support of NSF and the AO foundation, we have developed some novel techniques to address the following challenging problems whose solution is a prerequisite to semantic video retrieval: (a) A novel semantic-sensitive video content representation framework to enhance the quality of features and thus make the semantic gap bridgeable; (b) A new technique to detect the elementary units that are relevant to semantic medical concepts under different vision purposes; (c) A semantic medical concept interpretation model via flexible mixture model; (d) An adaptive Expectation-Maximization (EM) algorithm for model selection and parameter estimation; (e) An integrated EM algorithm to integrate unlabeled samples for classifier training and to address concept drift in the learning procedure. We have also demonstrated each of these techniques in implemented systems. Additional discussion can be found at: http://www.cs.uncc.edu/~jfan Jianping Fan got his MS degree in theoretical physics from Northwestern University and PhD in computer science and optical storage from the Chinese Academy of Sciences. When he was pursuing his PhD degree, he was one of the leaders in developing a new MPEG standard for compressing and storing videos more efficiently in compact discs (called Super-VCD) that is the only MPEG standard from a developing country. From 1997 to 1999, he was a JSPS professor in Osaka University, Japan. From 1999 to 2001, he was a visiting assistant professor at Purdue University and started his research on content-based video retrieval. From 2001, he has been an assistant professor at the University of North Carolina at Charlotte, where he has been building his own research area and group. He is the first author of more than 40 papers in top journals and leading conferences in image processing, coding and multimedia retrieval.