FINDING SONGS THAT SOUND THE SAME
Jean‐Julien Aucouturier, François Pachet · 2002
A crucial dimension of Content-based music management systems is their ability to compute automatically similarities be tween music titles. We propose a technique that allows users to find mu sic titles that sound similar to songs they like. The technique relies on a modelling of the timbral characteristics of a music s ignal by distributions of Cepstrum coefficients. The resulting model s are then compared to yield a similarity measure. The paper desc ribes the algorithm, and proposes an evaluation of the quality of t he extracted similarity measure. Additionally, we illustra te the use of this measure in two Electronic Music Distribution applica tions developed in the context of the European project Cuidado.