Music Genre Classification using CCN-based neuralnetworks

During the last decade, music streaming services have extended a lot and attracted many users around the world. One of themain challenges in the field of categorizing and recommending music to users is recognizing the genre of music. Music genre is aconventional category used to describe the characteristics of pieces of music that belong to a common tradition. Since genre is a high-levelattribute for a piece of music, it is a great step for correct classification. In this article, a method based on deep learning and ConvolutionalNeural Networks (CNN) is proposed, which performs genre classification with high accuracy. The proposed method consists of three mainphases: data pre-processing, training/verification phase and testing phase. The well-known GTZAN dataset has been used for evaluationof the proposed method, which has ۱۰۰۰ pieces of ۳۰ seconds music in ۱۰ different genres, some of which are: classical, blues, rock, hiphop,etc. Simulation results show that the proposed network can classify the data with an accuracy of ۸۰.۸%, which is better than that ofsome similar previous methods.

Keywords:

Genre classification , convolutional neural networks , deep learning , GTZAN