Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Applying statistical models and parametric distance measures for music similarity search

: Lukashevich, Hanna; Dittmar, Christian; Bastuck, Christoph


Fink, A.:
Advances in Data Analysis, Data Handling and Business Intelligence : Proceedings of the 32nd Annual Conference of the Gesellschaft für Klassifikation e.V., Joint Conference with the British Classification Society (BCS) and the Dutch/Flemish Classification Society (VOC), Helmut-Schmidt-University, Hamburg, July 16-18, 2008
Berlin: Springer, 2009 (Studies in classification, data analysis, and knowledge organization)
ISBN: 978-3-642-01043-9
ISBN: 978-3-642-01045-3
ISBN: 978-3-642-01044-6
Gesellschaft für Klassifikation (GfKl Annual Conference) <32, 2008, Hamburg>
Fraunhofer IDMT ()
music similarity; statistical models; gaussian mixture models; Kullback-Leibler divergence; music information retrieval; music similarity

Automatic deriving of similarity relations between music pieces is an inherent field of music information retrieval research. Due to the nearly unrestricted amount of musical data, the real-world similarity search algorithms have to be highly efficient and scalable. The possible solution is to represent each music excerpt with a statistical model (ex. Gaussian mixture model) and thus to reduce the computational costs by applying the parametric distance measures between the models. In this paper we discuss the combinations of applying different parametric modelling techniques and distance measures and weigh the benefits of each one against the others.