6533b826fe1ef96bd1284f6b

RESEARCH PRODUCT

A Fuzzy-Clustering Based Approach for Measuring Similarity Between Melodies

Brian MartínezVicente Liern

subject

MelodyFuzzy clusteringSimilarity (network science)Degree (graph theory)Computer sciencebusiness.industryFeature (machine learning)Pattern recognitionArtificial intelligenceCluster analysisbusinessFuzzy logicComplement (set theory)

description

Symbolic melodic similarity aims to evaluate the degree of likeness of two or more sequences of notes. In this work, we propose the use of fuzzy c-means clustering as a tool for the measurement of the similarity between two melodies with a different number of notes. Moreover, we present an algorithm, FOCM, implemented in a computer program written in C\(\sharp \) able to read two melodies from files with MusicXML format and to perform the clustering to calculate the dissimilarity between any two melodies. In addition, for each iteration step in the convergence process of the algorithm, a family of intermediate states (transition melodies) are obtained that can be used as new thematic material. This last feature, could be especially useful in the near future, as a complement in computer-aided composition.

yearjournalcountryeditionlanguage
2017-01-01
https://doi.org/10.1007/978-3-319-71827-9_21