Music Genre Identification

Abstract

In this project I developed a system based on machine learning that helps to recognize the genre of given music. The idea of this project came from IEEE paper that I have been read; B. Liang and M. Gu, "Music Genre Classification Using Transfer Learning," 2020 IEEE Conference on Multimedia Information Processing and Retrieval (MIPR). Music is an art that is widely celebrated worldwide, with countless songs being released or published every day. Music can be classified into many “genres”. Music genre is a classification of different instrumental material into conventional categories which helps in identification of a particular piece of music according to complementary cultural practices or set of cultural precedents. It is different from musical form and musical style, however, sometimes these terms are referred to interchangeably since there are fine distinctions between them. Genre is considered to be one of the most widely used measures to form separate clusters of music. In this paper, we focused on how a hierarchy of musical genres is important and also explore the importance of classifying audio files into these genres. After this, here define three feature sets namely timbral texture, rhythmic content and pitch content for the proposed classification task at hand. Then use the various statistical pattern recognition classifiers such as in order to classify full length as well as frame-based audio files. Here, we proposed a Convolutional Neural Network approach for audio-based classification. The task of this CNN involves using the spectrograms made out of the audio files to predict the 10 western music genres, including Rock, Reggae, Pop, Hip-hop, Disco, Country, Metal, Jazz, Blues and Classical Music.