Implementació d'un sistema no supervisat amb adaptació online per classificar música

Abstract

[en] Nowadays, music is still an important part of our lives: every one of us listens to some kind of music. There are currently loads of different styles and, to keep them organized, musical analysis has become very important - either to classify them according to their genre or for being able to recommend listeners similar songs to the ones they usually listen to. In this project we propose a unsupervised system to classify songs into different genres and, when necessary, label new genres that may come up. To achieve that, we will create groups from the Hierarchical clustering algorithm and, finally, classify and create new groups from the ECOC (Error-Correcting Output Codes) strategies.