KuralHubStar

KuralHubDatasetsTelugu

Telugu speech emotion recognition datasets

KuralHub lists 2 speech emotion recognition datasets for Telugu, 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. Telugu is also part of the benchmark. The best of 11 models reached 53.0% test accuracy (hubert-base-ls960).

Language family
Dravidian (South-Central Dravidian)
ISO code
te
Spoken mainly in
India
Datasets listed
2
In benchmark
Yes

Datasets at a glance

DatasetTypeSamplesSpeakersRecordingAccessLicence
IITKGP-SESCAudio1200010ActedOn request
Sowmith et al (2024)Audio458ActedOpenApache 2.0

Benchmark results

Telugu 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 40.5% (median 43.0%). The best run reached 53.0% with hubert-base-ls960. Compare with other languages.

Sowmith et al (2024) (run ID Telugu-Dataset). Scores on the test split, best run highlighted.
ModelFamilyVal acc.Test acc.Macro-F1Weighted-F1Logs
hubert-base-ls960bestHuBERT0.480.530.520.53logs
hubert-large-ls960-ftHuBERT0.410.350.300.30logs
wav2vec2-basewav2vec 2.00.460.440.420.43logs
wav2vec2-large-960hwav2vec 2.00.280.350.270.27logs
wav2vec2-large-lv60wav2vec 2.00.220.280.200.21logs
wav2vec2-xls-r-300mwav2vec 2.00.280.290.240.25logs
wav2vec2-xls-r-1bwav2vec 2.00.190.290.240.26logs
wavlm-base-plusWavLM0.440.530.500.52logs
wavlm-largeWavLM0.460.430.400.39logs
whisper-smallWhisper0.500.520.510.52logs
whisper-largeWhisper0.460.450.430.45logs

IITKGP-SESC

Indian Institute of Technology Kharagpur Simulated Emotion Speech Corpus

Type
Audio
Emotion labels
Happiness, Surprise, Anger, Disgust, Sadness, Fear, Sarcasm, Neutral
Size
samples - 12000, speakers - 10
Speakers
10
Recording
Acted
Access
On request
Licence
Not stated
File format
.wav
Website
Dataset Homepage
Paper
Paper Link
Contact
ksrao@iitkgp.ac.in

The IITKGP Simulated Emotion Speech Corpus (IITKGP-SESC) is a Telugu speech emotion dataset recorded using professional artists from All India Radio (AIR), Vijayawada, India. It contains emotionally simulated speech covering eight emotions: Happiness, Surprise, Anger, Disgust, Sadness, Fear, Sarcasm, and Neutral, derived from neutral statements. The dataset is designed for emotion analysis in speech, enabling the development of emotion-specific models using features from the vocal tract system, excitation source, and prosody. The quality of emotions is validated through subjective listening tests, and statistical models are used for classification.

Citation

@InProceedings{10.1007/978-3-642-03547-0_46,
  author="Koolagudi, Shashidhar G.
  and Maity, Sudhamay
  and Kumar, Vuppala Anil
  and Chakrabarti, Saswat
  and Rao, K. Sreenivasa",
  editor="Ranka, Sanjay
  and Aluru, Srinivas
  and Buyya, Rajkumar
  and Chung, Yeh-Ching
  and Dua, Sumeet
  and Grama, Ananth
  and Gupta, Sandeep K. S.
  and Kumar, Rajeev
  and Phoha, Vir V.",
  title="IITKGP-SESC: Speech Database for Emotion Analysis",
  booktitle="Contemporary Computing",
  year="2009",
  publisher="Springer Berlin Heidelberg",
  address="Berlin, Heidelberg",
  pages="485--492",
  abstract="In this paper, we are introducing the speech database for analyzing the emotions present in speech signals. The proposed database is recorded in Telugu language using the professional artists from All India Radio (AIR), Vijayawada, India. The speech corpus is collected by simulating eight different emotions using the neutral (emotion free) statements. The database is named as Indian Institute of Technology Kharagpur Simulated Emotion Speech Corpus (IITKGP-SESC). The proposed database will be useful for characterizing the emotions present in speech. Further, the emotion specific knowledge present in speech at different levels can be acquired by developing the emotion specific models using the features from vocal tract system, excitation source and prosody. This paper describes the design, acquisition, post processing and evaluation of the proposed speech database (IITKGP-SESC). The quality of the emotions present in the database is evaluated using subjective listening tests. Finally, statistical models are developed using prosodic features, and the discrimination of the emotions is carried out by performing the classification of emotions using the developed statistical models.",
  isbn="978-3-642-03547-0"
}

Source: datasets/Telugu/IITKGP-SESC.md · Suggest an edit

Sowmith et al (2024)

Type
Audio
Emotion labels
Happy, Sad, Angry, Neutral, Surprised
Size
samples-468
Recording
Acted
Access
Open
Licence
Apache 2.0
File format
.wav
Website
Dataset Homepage
Download
Download

The Telugu Emotion Speech dataset is a collection of 468 audio samples in the Telugu language, designed for speech emotion recognition. It includes five emotion categories: Happy, Sad, Angry, Neutral, and Surprised. This dataset is valuable for emotion analysis in Telugu speech and can be used for training and evaluating machine learning models in affective computing.

Used in the KuralHub benchmark (run ID Telugu-Dataset). See the results above.

Citation

@misc{sowmith2024telugu,
  author = {Jetty Sowmith and Others},
  title = {Telugu Emotion Speech},
  year = {2024},
  publisher = {Kaggle},
  howpublished = {\url{https://www.kaggle.com/datasets/jettysowmith/telugu-emotion-speech}},
  note = {Accessed: 2025-02-26}
}

Source: datasets/Telugu/Sowmith_et_al_2024.md · Suggest an edit


Know another Telugu speech emotion dataset, or a newer version of one listed here? Suggest it or open a pull request.

This page is generated from datasets/Telugu/ in the KuralHub repository. If you use a dataset, cite its authors. If KuralHub helped you find it, please also cite the KuralHub paper.