Classification of Music Genres using Multimodal Deep Learning Technique

¹ Gayatri Vidya Parishad College of Engineering for Women, Visakhapatnam
² CVR College of Engineering, Hyderabad
³ Sphoorthy Engineering College, Hyderabad

^* Corresponding author: janardhan.bitra@gmail.com

Abstract

The demand for automated music organization and the ever-increasing volume of digital audio recordings has both contributed to a surge in interest in deep learning-based genre classification. The purpose of this research is to examine the feasibility of using CNNs and RNNs, two types of deep learning architectures, for the task of audio track genre classification. The proposed models aim to achieve high accuracy and robustness in genre classification tasks by leveraging features extracted from raw audio signals and spectrogram representations. A comprehensive dataset comprising diverse music genres is utilized for training and evaluation, with performance metrics such as accuracy, precision, and recall assessed to ensure reliability. The results demonstrate that deep learning approaches significantly outperform traditional methods, providing insights into the underlying characteristics of musical styles. Potentially useful in areas such as music discovery platforms, playlist creation, and recommendation services, this study adds to the body of knowledge on automated music classification systems.