Generative Adversarial Networks Based Framework for Music Genre Classification

Music genre classification plays a crucial role in organizing and exploring large music collections, enabling personalized music recommendations, and enhancing music-related services. This paper presents a novel approach to music genre classification using Generative Adversarial Networks (GANs), Fourier Transform, and Wavelet Transform. The main objective is to leverage the power of GANs to extract discriminative features from audio data and accurately classify music into different genres. The proposed methodology involves two key components: the generator and the discriminator. The generator generates synthetic audio samples that resemble real music, while the discriminator learns to distinguish between real and synthetic audio samples. By training the GAN on a diverse dataset of music samples from various genres, the discriminator becomes proficient in recognizing genre-specific features. To enhance classification accuracy, Fourier Transform and Wavelet Transform are applied to extract both frequency and time-domain features from the audio data. Additionally, classifiers such as support vector machines and neural networks are employed to effectively distinguish between different music genres. The experimental results demonstrate the effectiveness of the proposed approach across multiple datasets. The method achieves 98.97% accuracy on the GTZAN dataset, 92.47% accuracy on the FMA-Small dataset, and 92.98% accuracy on the ISMIR Genre dataset, significantly outperforming traditional classification methods These results highlight the power of GANs, Fourier Transform, and Wavelet Transform in enhancing the accuracy and robustness of music genre classification. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2024.