Music Genre Classification Based on Machine Learning Methods

Abstract

A large portion of the population globally are actively downloading and streaming music from online platforms in recent years. In many parts of the world, music has become a lifestyle instead of a luxury. As a result of music’s growing popularity, the number of sounds and music released each day and also increased tremendously as the demand growth of music. Music genre is defined as a label that is descriptive of the music category which is used to categorize music based on several characteristics including: harmonic contents, pitch, instrumentation, and rhythmic structure. A classification model used to complete this task of classifying music genres is a machine learning system designed to classify using audio signals from song tracks into different musical genres. This paper aims to investigate the different approaches, including k-nearest neighbours algorithm, random forest, artificial neural network, and convolutional neural network, to classify music genres. Model performance analysis and confusion matrix analysis are applied to compare the advantages of the different algorithms applied on music genre classification problem.