Genre Classification of Music Using Machine Learning

Abstract

One of the most important technological advances of the 21st century is artificial intelligence (AI) and machine learning. They are revolutionizing computers, banking, healthcare, agriculture, music and travel. A strong model has mastered many difficult learning tasks. Speech analysis is one area of artificial intelligence. This includes finding information about music, creating music, and categorizing music. Music data is one of the most complex source data available today. This is primarily because it is challenging to pull relevant correlation data from it. Various methods, from classical to neural networks to hybrids, have been tested on music data and achieved excellent accuracy. This study intends to analyse and contrast various approaches to identify the musical genres. On a small sample of the Free Music Archive (FMA) dataset, the accuracy rates were as follows: 46% using Support Vector Classifier (SVC), 40% using Logistic Regression, 67% using Artificial Neural Network (ANN), 77% using Convolutional Neural Networks (CNN), 90% using Convolution-Recurrent Neural Network (CRNN), 88% using Parallel Convolution-Recurrent Neural Network (PCRNN), 73% without using Ensemble technique, and 85% using Ensemble technique of AdaBoost. We established SVC as our baseline model, which had an accuracy of 46%, and we defined the succeeding models to have an accuracy higher than that. For the test dataset, ANN provided us with a score of 67%, whereas CNN outperformed it with a score of 77%. We discovered that imagebased features performed better at categorising the labels than typical audio-extracted features. The dataset responded best to a mix of CNN and RNN, with a series CRNN model providing the highest accuracy. After that, we attempted to train an ensemble model to our dataset and investigated how it functioned. The numerous ways for classifying music genres are thoroughly examined in this research, with an emphasis on some parallel models and ensembling strategies.