AN ANALYSIS OF THE MONGEAU-SANKOFF ALGORITHM FOR MUSIC INFORMATION RETRIEVAL

Abstract

An essential problem in music information retrieval is to determinethesimilaritybetweentwogivenmelodies; there are several melodic similarity measures that have been proposed, among others, the Mongeau-Sankoff measure. In this work we implemented a modified version of the Mongeau-Sankoff measure. We conducted an experimental study to compare the implemented measure with other similarity measures; this evaluation was done in the context of the 2005 edition of the MIREX symbolic melodic similarity competition. The most relevant result of our work is an implementation of the Mongeau-Sankoff measure that presents greater effectiveness when compared to other current melodic similarity measures.