Deep Learning-Based Music Genre Classification using Convolutional Neural Network

Humans and other living beings express their emotions through music, making the classification of music genres essential. Music genre classification involves categorizing music items into predefined genres, such as classical, rock, and others. To develop a music genre classification system, we utilize neural networks with the GTZAN dataset, which comprises 10 genres, each containing 100 audio files that are 30 seconds long. Spectrogram images derived from audio signals serve as inputs for the neural network model to classify these genres. This research employs the Mel-Frequency Cepstral Coefficient (MFCC) technique to extract features from the audio signals. The raw MFCC matrix, consisting of numerical values, is used to train the model. Various models are constructed using both neural networks and convolutional neural networks, and their performances are compared. The experimental results demonstrate that the CNN model achieves superior prediction accuracy of $\mathbf{9 8 \%}$.