Paper Contents
Abstract
The proliferation of digital music necessitates efficient automated methods for organization and retrieval. Music genre classification, a fundamental task in Music Information Retrieval (MIR), plays a crucial role in this context. Traditional methods often rely on handcrafted features and classical machine learning algorithms, which may not fully capture the complex timbral and temporal characteristics of music. This paper presents an automated music genre classification system employing Deep Convolutional Neural Networks (CNNs). Audio signals are transformed into Mel-spectrograms, providing a rich time-frequency representation suitable for CNN-based feature learning. The proposed CNN architecture is designed to learn hierarchical discriminative features directly from these spectrograms. We evaluate our approach on the widely-used GTZAN dataset, demonstrating its effectiveness in accurately classifying music into ten distinct genres. The results indicate that CNNs offer a promising direction for robust music genre classification, achieving competitive performance and providing a scalable solution for large music repositories.
Copyright
Copyright © 2025 Sharan L. This is an open access article distributed under the Creative Commons Attribution License.
