StressDat

Database of  Speech Under Stress

On this page, we provide the StressDat – Speech Database under stress. Database includes 29 speakers (15 females, 14 males). They recorded pre-prepared utterances from 19 specific situations in different levels of stress . To ask for the StressDat database just fill out the EULA  and send the signed application to us by mail.

Form to send as plain text (in the body of your email) along with the EULA:

  • First name:
  • Last name:
  • Institution:
  • Country:
  • Institutional profile of the person signed in EULA:

Environment: Recordings were performed at home of each speaker. They were instructed to select a room with the smallest possible reverberations, for example having as few bare walls and surfaces as possible, and make adjustments to further improve the acoustic environment, such as spreading the curtains, opening wardrobes, or covering sharp furniture edges.

Audio quality:
Sampling rate: 16 000 Hz
Bitrate: 16 bit,
Channels: mono
Audio file format: Waveform Audio File Format (wav)
At the very beginning, every speaker recorded a sample, whose audio quality was checked.. In case of any problems with low/destroyed signal, reverberations etc. Speaker was asked for improvements or the recording was stopped.
Recoding device: Own smartphone of each speaker using default sound recorder at highest possible quality. Because of different quality of recordings they were downsampled to 16kHz,16bit,mono.

Text prompts:

From 10 to 13 pre-prepared sentences in 19 situations (12 stressful situations, 4 neutral situations, 3 soothing situations) in Slovak language. You can view the text prompts here.

Annotations

Each utterance was annotated by 5 annotators using “Stress Thermometer” based on the Subjective Units of Distress Scale.

Referencing:

All documents and papers that report on research that uses the StressDat database will acknowledge this by citing one of the following papers:

Sabo, R., Beňuš, Š., Trnka, M., Ritomský, M., Rusko, M., Schaper, M., & Szabo, J. (2021). StressDat–Database of speech under stress in Slovak. Jazykovedny Casopis, 72(2), 579-589.

Sample recordings:

Speaker nr. 23 (female), neutral: sample

Speaker nr. 23 (female), medium stress: sample

Speaker nr. 23 (female), high stress: sample

Speaker nr. 23 (female), shooting: sample