An automatic thaat and raga identification system using CNN-based models

Surya Majumder and Adrija Bhattacharya

Published in Innovations in Systems and Software Engineering (Oct 2023), Springer

Link to Paper

Link to Dataset (Thaat and Raga Forest (TRF) Dataset)

Abstract

Western music has been under the radar of computational musicology for quite some time now. However, Indian Classical Music is one of the prominent areas that is not extensively researched. In this manuscript, Deep Learning models are used in Indian Classical Music. It is often important to identify the Thaats and Ragas of a song computationally for further applications such as sentiment-based recommendation and identification. The Deep Learning model implemented here analyzes the Mel Spectrogram, STFT Spectrogram, and Chromagram of any song/music piece to classify it into different required categories with the help of CNN. On extensive testing, it is inferred that the proposed system is highly effective in the identification with high accuracy.

Dataset

Data Sample

Citation

If you're using this article or code in your research or applications, please consider citing using this BibTeX:

@Article{Majumder2023,
    author={Majumder, Surya and Bhattacharya, Adrija},
    title={An automatic thaat and raga identification system using CNN-based models},
    journal={Innovations in Systems and Software Engineering},
    year={2023},
    month={Oct},
    day={17},
    abstract={Western music has been under the radar of computational musicology for quite some time now. However, Indian Classical Music is one of the prominent areas that is not extensively researched. In this manuscript, Deep Learning models are used in Indian Classical Music. It is often important to identify the Thaats and Ragas of a song computationally for further applications such as sentiment-based recommendation and identification. The Deep Learning model implemented here analyzes the Mel Spectrogram, STFT Spectrogram, and Chromagram of any song/music piece to classify it into different required categories with the help of CNN. On extensive testing, it is inferred that the proposed system is highly effective in the identification with high accuracy.},
    issn={1614-5054},
    doi={10.1007/s11334-023-00541-2},
    url={https://doi.org/10.1007/s11334-023-00541-2}
    pages={1--16},
    year={2023},
    publisher={Springer}
}

Also if you are using our dataset in your research or applications, please consider citing using below BibTeX:

@data{0xtb-fn23-24,
    doi = {10.21227/0xtb-fn23},
    url = {https://dx.doi.org/10.21227/0xtb-fn23},
    author = {Majumder, Surya and Bhattacharya, Adrija},
    publisher = {IEEE Dataport},
    title = {Thaat and Raga Forest (TRF) Dataset},
    year = {2024}
}