Classification of Music Genre Audio Signals Using Deep Neural Networks

Abstract

Music is one of the fundamental parts in our lives. Music gives us amusement and relax when we feel empty. Everyone in the society will generally listen music of specific genre subject to how they are feeling right now. It turns into an exceptionally feverish and complex undertaking to classify the music manually. Thus, there is a requirement for a model to classify the music into various Genres which can be utilized in some genuine applications. In this task, a deep learning model is proposed to arrange the music into various types. The proposed model depends on the Convolutional Neural Networks, Which are generally excellent at image pattern recognition. The model is prepared over GZTAN dataset which contains ten classes with hundred music records connected with every sort. The music documents are first changed over into Mel-spectrograms utilizing librosa and afterward those spectrograms are fed as input to the developed Convolutional Neural Network model for preparing. So that the network extracts some features based on audio genre frequency and learns about different genres that are helpful to classify the unknown music file into respective genre.