| The storage and processing of non-textual data in computers, such as video, audio, and images, commonly referred to as multimedia data, have grown tremendously in recent times and are expected to have explosive growth in the coming years. This is due to the great need for multimedia data on the one hand, and the rapid growth in computational power in personal computers, the development of high-speed data communication networks, and the availability of high-density storage devices, on the other hand. Multimedia data can provide more effective dissemination of information and communication of ideas in science, engineering, medicine, education, and commerce.; Several inherent characteristics of multimedia data such as their huge sizes, unstructured and inexact nature, temporal dimension, coupled with the phenomenal increase in their usage in several applications have rendered traditional databases inadequate and necessitated the design of multimedia databases which have several new and extended features. Search for an object in a database is done using an index which is a feature characterizing the object. The huge sizes and inexact nature of multimedia data have rendered keyword-based indices and exact searches used in traditional databases ineffective, and have called for content-based indices and similarity searches in multimedia databases.; The enormous quantities of digital audio data being generated, processed, and consumed by users and applications, have created a critical need for the management of audio data using an audio database system. In recent efforts, audio data has not received a similar level of attention as image and video databases. In the context of the design of an audio database, the contributions of this thesis are: (1) the study and development of new indexing techniques to facilitate content-based queries, (2) organization of the index data in a suitable data structure to support efficient searches, and (3) development and analysis of efficient similarity-search schemes. These three components are keys to providing sophisticated, accurate and fast retrievals for queries in audio databases. |
