Abstract
Music genre identification is crucial for the classification and recommendation of songs in music applications. Manually labelling songs takes up a significant amount of time. In this paper, we propose a deep learning model to automate the process of genre identification. The process mainly involves three steps: preprocessing the dataset to get a simplified version of each song, building a deep neural network, and training and using it to predict the genre of songs. Input to the model is Mel-frequency Cepstral Coefficient (MFCC) values of the audio files from the GTZAN dataset that consists of 10 different genres. After training, the model produced a result of 60% accuracy. Observing the actual and predicted values, the model seemed to exhibit overfitting. To overcome this, we used dropouts and regularization in the model, followed by early stopping, which gave a final accuracy of 67.5%.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call