Improvements in technology have increased the amount of human interactive systems that support visual and audial operations. Besides many others, especially the recent entertainment industry has been built on the digital world and processing large collections of multimedia materials. Having huge amount multimedia data revealed the need for efficient and effective ways of modeling, storing, addressing, and retrieving such huge data, mostly, the semantic contents in it. Although there are some database management systems that support multimedia objects by some add-ons or extensions, they are far away from satisfying user requests, which are usually based on similarity and semantic contents. In this study, the requirements of multimedia databases are analyzed. To satisfy such requirements, a database architecture which is specialized for multimedia objects is constructed and a conceptual data model, handling semantic multimedia contents is implemented. In the architecture, a semantic information extractor, a high-dimensional index structure and an intelligent fuzzy object-oriented database component are integrated through a coordinator structure. The proposed architecture is realized and a prototype system is implemented by tightly coupling several multimedia modules.