Dutch speech emotion recognition datasets

KuralHub lists 1 speech emotion recognition dataset for Dutch. Each entry below gives the emotion labels, size, recording setup, licence, where to get it and how to cite it.

Language family: Indo-European (Germanic)
ISO code: nl
Spoken mainly in: Netherlands, Belgium, Suriname
Datasets listed: 1
In benchmark: No

Datasets at a glance

Dataset	Type	Samples	Speakers	Recording	Access	Licence
PICKA-NH	Audio		113	Studio	Open	MIT, CC-BY

PICKA-NH

Type: Audio
Emotion labels: Happiness, Anger, Sadness
Speakers: 113
Recording: Studio
Access: Open
Licence: MIT, CC-BY
Website: Dataset Repository
Paper: Development of vocal emotion recognition in school-age children: The EmoHI test for hearing-impaired populations
Download: Request
Contact: mail

The dataset comprises recordings from six native Dutch speakers (three male, three female) who articulated two pseudospeech sentences—"Koun se mina lod belam" and "Nekal ibam soud molen"—expressing three core emotions: happy, sad, and angry. These sentences are non-meaningful in Dutch, English, or any other Indo-European languages, ensuring a focus on emotional prosody without semantic influence. Recordings were conducted in an anechoic chamber at a 44.1 kHz sampling rate.

An online survey with Dutch and English adults was utilized to evaluate the recognizability of these emotional expressions, leading to the selection of four speakers whose renditions were most accurately identified. The final stimulus set includes 36 experimental stimuli, with each speaker contributing three productions per emotion, and an additional four practice stimuli for training purposes.

Participants consisted of 58 Dutch and 25 English children aged 4 to 12, and 15 Dutch and 15 English adults aged 20 to 30, all with normal hearing and monolingual language backgrounds. Additionally, preliminary data were collected from 14 prelingually deaf Dutch children with cochlear implants aged 4 to 16. Normal hearing participants underwent pure-tone audiometry to confirm hearing thresholds at 20 dB HL across octave frequencies between 500 and 4,000 Hz. Written informed consent was obtained from adult participants and the parents or legal guardians of child participants prior to data collection.

Testing was conducted in quiet environments: normal hearing children were assessed at their homes, while adults were tested in university settings. The experiment commenced with a training session of four practice stimuli, followed by 36 experimental stimuli presented in a randomized order, lasting approximately 6 to 8 minutes. Stimuli were delivered via Sennheiser HD 380 Pro headphones for normal hearing participants and Logitech Z200 loudspeakers for cochlear implant children, calibrated to 65 dBA. Cochlear implant children used their regular device settings consistently throughout the experiment. Participants responded by selecting one of three clowns on a touchscreen interface corresponding to the perceived emotion, receiving visual feedback—confetti for correct responses and a head-shaking parrot for incorrect ones—to maintain engagement and indicate progress.

Citation

@article{nagels2020development,
  title={Development of vocal emotion recognition in school-age children: The EmoHI test for hearing-impaired populations},
  author={Nagels, Leanne and Gaudrain, Etienne and Vickers, Deborah and Lopes, Marta Matos and Hendriks, Petra and Ba{\c{s}}kent, Deniz},
  journal={PeerJ},
  volume={8},
  pages={e8773},
  year={2020},
  publisher={PeerJ Inc.}
}

Source: datasets/Dutch/PICKA-NH.md · Suggest an edit

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

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