Information Retrieval Using Markov Model Mediators in Multimedia Database Systems

Recent progress in high-speed communication networks, large capacity storage devices, digitalized media, and data compression technologies have resulted in a variety of multimedia applications using the integration of text, images, audio, graphics, animation, and full-motion video. For traditional text-based database management systems, data access and manipulation have advanced considerably. However, for multimedia database systems, information retrieval is more diicult than that of the conventional data since it is necessary to incorporate diverse media with diverse characteristics. The need for information retrieval in multimedia database systems increases proportional to the continuous growth of diverse information sources and the proliferation of independent but related user applications. Therefore, the ability to query the databases and to locate speciic information directly as needed is important for multimedia database systems. For this purpose, a mathematical sound framework, called Markov model mediators (MMMs) which employ the principle of Markov models and the concept of mediators, is introduced in this paper. The proposed MMM mechanism performs information retrieval via a stochastic process which generates a list of possible state sequences with respect to a given query and indicates which particular media objects to query.