A multimedia database management system supporting contents search in media data / [by] Vincent Y. Lum [and] Klaus Meyer-Wegener.

Abstract

It is now feasible to store and manage in computers new types of data like text, images, graphics, and sound recordings. This paper claims that database management systems should be extended to organize these new types of data and to enable search based on their contents. Media objects are modelled as attributes of abstract data types. The contents is captured in terms of natural language descriptions and is translated by a parser into predicates for easy matching with query phrases. The implications of this approach are discussed: Data organization for multimedia including contents representation, the dictionary used by the parser to recognize words and assign predicates, rules to use semantic relationships in the query evaluation, and access paths to speed up the search for the descriptions. The result is an architecture for multimedia database management systems that combines the additional components needed with the conventional data management and identifies their ways of interaction. Three types of user interfaces are offered that require different levels of skill. The architecture is modular and allows the integration of more advanced AI techniques once they become available