Central Concerns and Questions
The objective of the research is to develop effective and efficient data analysis and retrieval techniques for various novel rich/multimedia application domains. The research challenges and questions include
- how to combine diverse knowledge to achieve effective information access
- how to improve efficiency and scalability of the related techniques
- how to assist general users to express their information needs
- economic and robust system performance evaluation process, and;
- deep understanding of business, cultural and social impacts of rich media
Emerging Ideas and Initiatives
The proliferation of various rich/multimedia information and associated applications over the past decade has led to an increasingly colorful lifestyle. Having been applied for wide range of purposes such as entertainment, education, psychology, management and economy, unique characteristics of rich/multimedia data pose huge challenge for informative retrieval, knowledge discovery and data management. Distinguished from standard alphanumeric data, rich/multimedia information has much more complex structure which might involve spatial and/or temporal dependency. Also, from representation point of view, rich/multimedia data could be huge and contain rich high level semantic meaning in general. Consequently, to efficiently manage and assess multimedia information under a real life application environment, there is an urgent need for technological advances in data structure for efficient organisation, intelligent content representation and system performance evaluation. In this research, the problems of rich/multimedia information retrieval and data management have been addressed in three aspects:
- A novel indexing framework for large multimedia databases with superior scalability and effectiveness [1][2].
- Fast machine learning based methods to generate small and effective media object signature [3][4].
- Statistical based approach to derive accurate estimations on performance evaluation results [5].
Selected Publications
[1] Jialie Shen, John Shepherd, Bin Cui, Kian-Lee Tan. A Novel Framework for Efficient Automated Singer Identification in Large Music Databases. ACM Transactions on Information Systems (ACM TOIS), 27(3): 1-31, 2009.
[2] Jialie Shen. Stochastic Modeling Western Paintings for Effective Classification. Pattern Recognition, 42(3): 293-301, 2009.
[3] Bingjun Zhang, Jialie Shen, Qiaoliang Xiang, Ye Wang. CompositeMap: A Novel Framework for Music Similarity Measure. ACM SIGIR Conference on Research and Development in Information Retrieval (ACM SIGIR), July 2009.
[4] Jialie Shen, Dacheng Tao, Xuelong Li. Modality Mixture Projections for Semantic Video Event Detection. IEEE Transactions on Circuits and Systems for Video Technology, 18(11): 1587-1596, 2008.
[5] Jialie Shen, John Shepherd. Efficient Benchmarking of Content-based Image Retrieval via Resampling. ACM International Conference on Multimedia (ACM MM), 569-578, October 2006.
External Collaborations and Industry Linkages
- Christos Faloutsos, Carnegie Mellon University
- Ling Liu, Georgia Institute of Technology
- Calton Pu, Georgia Institute of Technology
- Yong Rui, Microsoft Research
- Anindya Ghose, Leonard N. Stern School of Business, NYU
