Modeling of Audio Data for Multi-Criteria Query Formulation

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


The amount of available audio data in multimedia databases is increasing rapidly in consequence of advancements in media creation, storage and compression technologies. This rapid increase imposes new demands in audio data management and retrieval. As a result, growing number of research works have been put into the area, such as audio indexing, classification and segmentation. However, modeling and retrieval techniques are not adequate and handling audio data content is still far from sufficient for most retrieval tasks. This work proposes an audio data and audio repository model to fulfill user requirements in retrieving audio data from large collections. The audio data repository model we proposed in this thesis enable us to capture the audio itself, its low-level representation, all alphanumeric data associated to the audio as well as timing information. Thus, it facilitates both keyword-based and similarity-based operations on audio objects. In the proposed model, a generic audio repository model that can handle a general audio as well as a sub-repository model (speech repository model) that can manipulate speech through its constituent units is discussed. The speech repository model enable us to capture all relevant information associated to speech units, to keep track of hierarchical relationships between speech units and the speech that contains them. The Object Relational scheme is used to manage these audio related data under a DBMS. The proposed work augments audio retrieval by enabling users to apply semantic concepts (highlevel features) linked to audio signals in their query in order to match relevant audio in a database. Such an approach improves simple matching based on audio signal characteristics (lowlevel features), as the user need not have example sounds for querying. In addition, an example audio can be used as a query since users may not always know exact search terms to achieve useful results. Finally, the practicality of the proposed model is demonstrated by taking sample application areas from the medical domain. Keywords: low-level features, high-level features, audio data model, audio repository model, content-based audio retrieval, keyword-based audio retrieval, Query-By-Example (QBE), ADMMA.



Low-Level Features, High-Level Features, Audio Data Model, Audio Repository Model, Content-Based Audio Retrieval, Keyword-Based Audio Retrieval, Query-By-Example (Qbe), Admma