DEEP LEARNING FOR CARNATIC AND NON-CARNATIC MUSIC CLASSIFICATION: A COMPARATIVE STUDY OF CNN AND RNN ARCHITECTURES

Chitrarasu Manikandan

doi:10.51278/ajse.v3i2.1645

Authors

Chitrarasu Manikandan AMET UNIVERSITY

DOI:

https://doi.org/10.51278/ajse.v3i2.1645

Keywords:

Carnatic Music Classification, Deep Learning Models, Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Audio Feature Extraction, Music Genre Analysis, Spectrogram, Mel-Frequency Cepstral Coefficients (MFCCs).

Abstract

This study articulates a deep learning approach for classifying Carnatic and Non-Carnatic
music using Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN). The
differential features, such as microtones, and improvisational structures of Carnatic music cause
severe difficulties in the application of automated genre classification. Thus, audio features of
MFCCs, chroma features, and spectrograms were extracted to capture key spectral and tonal
properties in order to realize excellent classification. The CNN model achieved an accuracy of 95.1%,
outperforming the RNN model's 93.8%, with ROC-AUC scores of 0.96 and 0.94, respectively. These
metrics indicate the CNN’s effectiveness in handling complex spatial features in audio data, while the
RNN provided valuable insights into sequential patterns. This Result highlights CNN’s advantages in
capturing the intricacies of genre classification for culturally rich music forms like Carnatic. Future
research will focus on increasing the performance of such models, and leveraging both spatial and
temporal dimensions in audio might happen using hybrid CNN-RNN architectures. Also, this research
contributes to the advancement of technology around music classification with promising avenues to
cultural preservation and digital archiving of music.

References

Q. G. Rafi, M. Noman, S. Z. Prodhan, S. Alam, and D. Nandi, “Comparative Analysis of Three Improved Deep Learning Architectures for Music Genre Classification,” Int. J. Inf. Technol. Comput. Sci., vol. 13, no. 2, pp. 1–14, 2021, doi: 10.5815/ijitcs.2021.02.01.

Z. Wang, C., Zhang, Y., Ding, H., “Music genre classification using deep learning: a comparative analysis of CNNs and RNNs,” Appl. Math. Nonlinear Sci., vol. 8, no. 2, pp. 3383–3392, 2023.

Y. H. Cheng and C. N. Kuo, “Machine Learning for Music Genre Classification Using Visual Mel Spectrum,” Mathematics, vol. 10, no. 23, 2022, doi: 10.3390/math10234427.

X. Han, W. Chen, and C. Zhou, “Musical Genre Classification Based on Deep Residual Auto-Encoder and Support Vector Machine,” J. Inf. Process. Syst., vol. 20, no. 1, pp. 13–23, 2024, doi: 10.3745/JIPS.04.0300.

L. Liu, “The implementation of a proposed deep-learning algorithm to classify music genres,” Open Comput. Sci., vol. 14, no. 1, 2024, doi: 10.1515/comp-2023-0106.

W.-H. Hsu, B.-Y. Chen, and Y.-H. Yang, “Deep Learning Based EDM Subgenre Classification using Mel-Spectrogram and Tempogram Features,” pp. 1–6, 2021, [Online]. Available: http://arxiv.org/abs/2110.08862.

X. He and F. Dong, “A deep learning-based mathematical modeling strategy for classifying musical genres in musical industry,” Nonlinear Eng., vol. 12, no. 1, 2023, doi: 10.1515/nleng-2022-0302.

M. Ashraf et al., “A Hybrid CNN and RNN Variant Model for Music Classification,” Appl. Sci., vol. 13, no. 3, 2023, doi: 10.3390/app13031476.

J. Salamon, C. Jacoby, and J. P. Bello, “A dataset and taxonomy for urban sound research,” MM 2014 - Proc. 2014 ACM Conf. Multimed., no. October, pp. 1041–1044, 2014, doi: 10.1145/2647868.2655045.

B. McFee et al., “Librosa: Audio and Music Signal Analysis in Python, In the Proceedings of the 14th Python in Science Conference, Austin, Texas, 6 - 12 July 2014,” Scipy, no. Scipy, pp. 18–24, 2015.

S. K. Tom Ko, Vijayaditya Peddinti, Daniel Povey, “Audio Augmentation for Speech Recognition,” Neurocomputing, vol. 100, pp. 144–152, 2013, [Online]. Available: http://dx.doi.org/10.1016/j.neucom.2011.09.037.

J. Schlüter and T. Grill, “Exploring data augmentation for improved singing voice detection with neural networks,” Proc. 16th Int. Soc. Music Inf. Retr. Conf. ISMIR 2015, pp. 121–126, 2015.

R. Singhal, S. Srivatsan, and P. Panda, “Classification of Music Genres using Feature Selection and Hyperparameter Tuning,” J. Artif. Intell. Capsul. Networks, vol. 4, no. 3, pp. 167–178, 2022, doi: 10.36548/jaicn.2022.3.003.

S. Oramas, F. Barbieri, O. Nieto, and X. Serra, “Multimodal Deep Learning for Music Genre Classification,” pp. 1–18, 2018, [Online]. Available: https://doi.org/xx.xxxx/xxxx.xx.

Navin Kamuni, “Enhancing Music Genre Classification through Multi-Algorithm Analysis and User-Friendly Visualization,” J. Electr. Syst., vol. 20, no. 6s, pp. 2274–2281, 2024, doi: 10.52783/jes.3178.

Q. Lemaire and A. Holzapfel, “Temporal convolutional networks for speech and music detection in radio broadcast,” Proc. 20th Int. Soc. Music Inf. Retr. Conf. ISMIR 2019, pp. 229–236, 2019.

W. Zhang, “Music Genre Classification Based on Deep Learning,” Mob. Inf. Syst., vol. 2022, 2022, doi: 10.1155/2022/2376888.

X. Pons, Jordi; Serra, “DESIGNING EFFICIENT ARCHITECTURES FOR MODELING TEMPORAL FEATURES WITH CONVOLUTIONAL NEURAL NETWORKS Jordi Pons and Xavier Serra Music Technology Group , Universitat Pompeu Fabra , Barcelona,” IEEE Int. Conf. Acoust. Speech, Signal Process. 2017, pp. 2472–2476, 2017.

M. Ge, C. Delgado-Battenfeld, and D. Jannach, “Beyond accuracy: Evaluating recommender systems by coverage and serendipity,” RecSys’10 - Proc. 4th ACM Conf. Recomm. Syst., no. April 2016, pp. 257–260, 2010, doi: 10.1145/1864708.1864761.