Bangla Music Genre Classification Using Neural Network

Abstract

Music genre classification is very vital for music recommendation and for the retrieval of music information. So many works have already been done for classifying genres of English music using different machine learning approaches. Even though Bangla music is very rich in its own fashion, there is almost no notable work found to classify music genres of Bangla music using machine learning techniques yet. There are so many types and styles of Bangla music which can be classified in different genres. Initially, we're considering 6 different Bangla music genres such as `Bangla Adhunik', `Bangla Hip-Hop', `Bangla Band Music', `Nazrulgeeti', `Palligeeti', `Rabindra Sangeet' etc. We are using 250-300 songs (. MP3 files) for each genre. We extracted different time domain and frequency domain features of audio signals from digital audio files (i.e. MP3 files). Finally, in this study, we proposed a deep learning model (after comparing performances of different models) to do a multiclass classification of Bangla music genres.