SURVEY on Music Genre Recognition using Deep Learning

Abstract

Music genre is clear cut names made by people to classes bits of music. A music genre classification is portrayed
by the basic attributes shared by its individuals. These attributes ordinarily are identified with the instrumentation, harmonic
content, and rhythmic structure of the music. Genre is usually used to structure the expansive accumulations of music
accessible on the Web. Presently, music genre annotation is performed manually. Automatic music genre classification
arrangement can help or supplant the human client in this procedure and would be a valuable expansion to music data
retrieval frameworks. Likewise, Automatic music genre classification gives a structure to creating and assessing highlights
for a substance based examination of musical signals. In this survey paper, the automatic music genre into a progressive
system of music genre is investigated. All the more particularly, three feature sets for representing pitch content, rhythmic
content and timbral texture are proposed. The performance and relative importance of the proposed features are investigated
by training statistical pattern recognition classifiers using real-world audio collections. We compare the classification
accuracy rate of various deep learning models with a set of well-known learning models including Deep Neural Network,
Convolution Neural Network, Recurrent Neural Network in combination with hand-crafted audio features for a genre
classification task on a public dataset.