You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

Dysarthric Speech (German)

A German dysarthric speech dataset of a single speaker with a speech impairment. The dataset is intended for research on automatic speech recognition (ASR), speech processing, and assistive technologies for speakers with dysarthria.

Access is gated. Please request access via the button above. Access is granted at the discretion of the dataset maintainer.

Dataset Structure

The dataset is organized by speaker. Currently one speaker is included:

Doehring/

Two recording types are provided:

Type Description Audio files Utterances
Read speech Prompted read-aloud recordings 109 MP3s 22,998 segments
Corrections Spontaneous correction recordings 4,446 MP3s 4,445 utterances

Splits (read speech)

Split Utterances
Train 22,572
Dev 209
Test 217

Files

File Description
impairedSpeech.DE.{train,dev,test}.cased Transcriptions, one sentence per line
impairedSpeech.DE.{train,dev,test}.seg.aligned Utterance ID, audio path, start time, end time (in centiseconds)
corrections.cased Transcriptions for correction recordings
corrections.seg.aligned Utterance ID and audio path for correction recordings
Doehring_readspeech/ Read speech MP3 audio files
Doehring_corrections/ Correction MP3 audio files

Language

German (de)

License

CC BY-NC 4.0 — free to use for non-commercial research with attribution.

Citation

If you use this dataset in your research, please contact the maintainer for the appropriate citation.

Downloads last month
2