Automatic Identification of Traditional Colombian Music Genres based on Audio Content Analysis and Machine Learning Technique

TL;DR

This paper proposes a machine learning-based method to automatically classify Colombian traditional music genres from audio content, achieving an average accuracy of 69% across six genres, facilitating large-scale music cataloging.

Contribution

It introduces a novel audio feature extraction and classification approach specifically tailored for Colombian music genres, addressing the challenge of automatic genre identification in complex rhythms.

Findings

Achieved 69% average accuracy in genre classification

Successfully distinguished six Colombian folkloric music genres

Demonstrated feasibility of automated music genre identification in complex rhythms

Abstract

Colombia has a diversity of genres in traditional music, which allows to express the richness of the Colombian culture according to the region. This musical diversity is the result of a mixture of African, native Indigenous, and European influences. Organizing large collections of songs is a time consuming task that requires that a human listens to fragments of audio to identify genre, singer, year, instruments and other relevant characteristics that allow to index the song dataset. This paper presents a method to automatically identify the genre of a Colombian song by means of its audio content. The method extracts audio features that are used to train a machine learning model that learns to classify the genre. The method was evaluated in a dataset of 180 musical pieces belonging to six folkloric Colombian music genres: Bambuco, Carranga, Cumbia, Joropo, Pasillo, and Vallenato. Results show that it is possible to automatically identify the music genre in spite of the complexity of Colombian rhythms reaching an average accuracy of 69\%.