KuralHub lists 3 speech emotion recognition datasets for Hindi, 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. Hindi is also part of the benchmark. The best of 11 models reached 71.0% test accuracy (whisper-small).
Hindi 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 48.8% (median 61.0%). The best run reached 71.0% with whisper-small. Compare with other languages.
VishalB (run ID Hindi-Dataset). Scores on the test split, best run highlighted.
The Emotional Hindi Speech Corpus developed at IPEC includes recordings from 28 speakers (14 male, 14 female) aged 20–30, expressing six emotions across 12 sentences. It is designed for applications in speech recognition, synthesis, speaker identification, and emotion analysis. The dataset also examines the impact of emotions on F0 contour and spectrum.
Citation
@INPROCEEDINGS{6709867,
author={Bansal, Sweeta and Dev, Amita},
booktitle={2013 International Conference Oriental COCOSDA held jointly with 2013 Conference on Asian Spoken Language Research and Evaluation (O-COCOSDA/CASLRE)},
title={Emotional hindi speech database},
year={2013},
volume={},
number={},
pages={1-4},
keywords={Speech;Databases;Speech recognition;Mobile communication;Spectrogram;Educational institutions;Time-frequency analysis;Speech corpus;Emotional speech;F0Contour},
doi={10.1109/ICSDA.2013.6709867}}
The dataset created by Koolagudi et al. in 2011 is an emotional speech corpus recorded in Hindi, featuring 10 speakers (5 male and 5 female) within the age ranges of 28–48 years for males and 20–30 years for females. Each speaker expressed 8 different emotions by speaking 15 sentences in a single session, with a total of 10 sessions conducted for corpus collection. The dataset comprises 12,000 utterances, amounting to 9 hours of recordings. The number of syllables per sentence ranges from 9 to 17, while the word count varies between 4 and 7. This dataset is valuable for emotional speech analysis and natural language processing applications.
Note: Fill the requested form and send it.
Citation
@INPROCEEDINGS{5738540,
author={Koolagudi, Shashidhar G. and Reddy, Ramu and Yadav, Jainath and Rao, K. Sreenivasa},
booktitle={2011 International Conference on Devices and Communications (ICDeCom)},
title={IITKGP-SEHSC : Hindi Speech Corpus for Emotion Analysis},
year={2011},
volume={},
number={},
pages={1-5},
keywords={Speech;Databases;Emotion recognition;Support vector machines;Feature extraction;Speech recognition;Computational modeling},
doi={10.1109/ICDECOM.2011.5738540}}
Know another Hindi speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.
This page is generated from datasets/Hindi/ in the KuralHub repository.
If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.