Music genres classification by deep learning

Abstract

Since musical genre is one of the most common ways used by people for managing digital music databases, music-genre-classification is a crucial task. There are many scenarios for its use, and the main one explored here is eventually being placed on Spotify, or Netease music, as an external component to recommend songs to users. This paper provides various deep neural networks developed based on python, together with the effect of these models on music genres classification. In addition, the paper illustrates the technologies for audio feature extraction in industrial environment by mel frequency cepstral coefficients (MFCC), audio data augmentation in order to analyze the music species based on local data feature comparison.