Music Genre Classification- A Holistic Approach Employing Multiple Features
K. Anuraj, S. S. Poorna, S Renjith · 2024
There is a significant increase in content related to digitized music, along with the rapid growth of digital multimedia. Hence the music genre categorization plays a crucial role in this digital era. It is useful for searching and retrieving user-specific information, as well as recommendation systems. This work is a comparative study on the performance of the combination of various Artificial Intelligence models as well as various features in music genre classification using GTZAN dataset. In order to carryout the pilot study, three types of features, viz. Mel spectrograms, Mel frequency cepstral coefficients and features based on Wave2Vec are employed. AI techniques encompassing deep learning and machine learning models are utilized to effectively classify the music types from the aforementioned features. Results of this study show that the combination of features based on Wave2Vec with 1D Convolutional Neural Network yielded the highest accuracy of 82%.