'A MUSIQUE Framework - Efficient and Extensible Framework for Audio Music Similarity Querying'
Authors
Schulz, Kim ; Jensen, Per Bech ; Rajaratnam, Kovarththanan
Term
10. term
Publication year
2006
Abstract
Vi præsenterer en fleksibel ramme til at bygge multimediesystemer med lighedssøgning oven på en relationsdatabase. Lighedssøgning betyder, at man kan finde billeder, lyd eller video, der minder om et givet eksempel, i stedet for kun at søge efter nøjagtige match. Rammen understøtter forskellige implementeringer af dens komponenter. Vi har særligt fokus på en indeksstruktur kaldet M-Grid, som kan integreres i rammen. Vi beskriver formelt M-Grid, gennemgår design, implementering og optimeringer og evaluerer deres effekt. Evalueringen viser, at rammen er fleksibel og robust og egner sig både til testsystemer og løsninger i praksis, hvilket understøttes af flere casestudier.
We present a flexible framework for building multimedia systems with similarity queries on top of a relational database. Similarity queries let users find images, audio, or video that resemble a given example, rather than exact matches. The framework supports different implementations of its components. We place special emphasis on an index structure called M-Grid, which can be integrated into the framework. We provide its formal description, design, implementation, and optimizations, and evaluate their impact. Our evaluation shows the framework is flexible and robust, making it suitable for both test systems and real-world deployments, as supported by several case studies.
[This abstract was generated with the help of AI]
Documents