Automatic classification of music genres based on constrained clustering

• Grant number: 17/07525-2
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): August 01, 2017
• Effective date (End): July 31, 2018
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Thiago Ferreira Covões
• Grantee: Estevão Crippa da Veiga
• Host Institution: Centro de Matemática, Computação e Cognição (CMCC). Universidade Federal do ABC (UFABC). Ministério da Educação (Brasil). Santo André , SP, Brazil

Abstract

Automatic classification of music genres is an important task for analyzing music data. One of the reasons that makes this task hard is the absence of precise definition about each genre. As a consequence, several taxonomies are adopted in different applications. In this work, we will assess the use of constrained clustering to identify clusters of musics. Such approach will make it possible to identify sub-genres without an a priori definition by the user. Evolutionary Algorithms will be used to perform the clustering such that an efficient search for Gaussian Mixture Models that best fit the data is performed. For assessing our approach, we will use benchmark datasets from the music information retrieval community. The candidate will learn and implement techniques related to Machine Learning and Evolutionary Algorithms. (AU)