Automatic Song Genre Classification in Bengali Music: A Comparative Study of Machine Learning and Deep Learning Approaches

Music genre categorization primarily refers to the identification of the type of music. Genre categorization is useful to create and organise our personalised playlist, including enhancing our musical experiences through music suggestions. Currently, for the purpose of categorising the genre of music, various machine learning and deep learning algorithms have been used, but it has been noticed that there is a lack of research in the classification of the BangIa music genre. Despite having some works in this field, the performance of the proposed models are not very efficient. Therefore, in this paper, we have used some leading-edge models of machine learning as well as deep learning to classify the genres of BangIa music. Six different BangIa music genres are represented in the dataset we use. Also, the dataset consists of different important features of music, such as spectral bandwidth, chroma frequency, spectral roll-off, zero crossing value, mfcc, etc. We went through intense data pre-processing, and with the assistance of a diverse range of metrics, the performance of our proposed models were assessed for multi-class classification. Moreover, it is worthy of consideration that our implemented deep neural network achieved an accuracy of about 83.65 percent.