KuralHub lists 2 speech emotion recognition datasets for Kannada, 1 of them openly available. Each entry below gives the emotion labels, size, recording setup, licence, where to get it and how to cite it. Kannada is also part of the benchmark. The best of 11 models reached 48.0% test accuracy (wavlm-large).
Kannada is one of the 29 languages in the KuralHub benchmark. Each of the 11 frozen speech models was trained with a small classification head on the shared emotions and scored on a held-out test split. Over all 11 runs the mean test accuracy was 27.5% (median 26.0%). The best run reached 48.0% with wavlm-large. Compare with other languages.
Agrawal et al (2022) (run ID Kannada-Dataset). Scores on the test split, best run highlighted.
I introduce a Kannada emotional speech dataset with 468 audio samples, contributed by 13 native speakers (4 male, 9 female). Each speaker recorded six sentences in five emotions plus neutral. Recordings were made in an uncontrolled environment. The dataset follows the naming convention AA-EE-SS.wav and is freely available under a Creative Commons license.
Used in the KuralHub benchmark (run ID Kannada-Dataset). See the results above.
Citation
@dataset{agrawal_2022_6345107,
author = {Agrawal, Vishakha},
title = {A Kannada Emotional Speech Dataset},
month = mar,
year = 2022,
publisher = {Zenodo},
doi = {10.5281/zenodo.6345107},
url = {https://doi.org/10.5281/zenodo.6345107},
}
@inproceedings{geethashree2017kannada,
title={Kannada emotional speech database: design, development and evaluation},
author={Geethashree, A and Ravi, DJ},
booktitle={Proceedings of International Conference on Cognition and Recognition: ICCR 2016},
pages={135--143},
year={2017},
organization={Springer}
}
Know another Kannada speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.
This page is generated from datasets/Kannada/ in the KuralHub repository.
If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.