The Ultrax 2020 Dataset - UX2020
A dataset of ultrasound and audio recorded with children with speech sound disorders
The Ultrax 2020 dataset is a collection of ultrasound tongue imaging and audio data, gathered from children with speech sound disorders by speech and language therapists in hospital environments.
Speakers
We recorded data with 43 English-speaking children, but only 37 gave consent to share their data. These are 11 female speakers and 26 male, aged 5-12 years.
| SPEAKER-ID | GENDER | AGE-Y | AGE-M | AGE |
|---|---|---|---|---|
| 01F | F | 7 | 6 | 7.50 |
| 02F | F | 10 | 1 | 10.08 |
| 03M | M | 5 | 7 | 5.58 |
| 04M | M | 7 | 11 | 7.92 |
| 05M | M | 10 | 2 | 10.17 |
| 06F | F | 10 | 5 | 10.42 |
| 07M | M | 10 | 4 | 10.33 |
| 08F | F | 7 | 6 | 7.50 |
| 09M | M | 10 | 11 | 10.92 |
| 10M | M | 7 | 4 | 7.33 |
| 11M | M | 10 | 11 | 10.92 |
| 12M | M | 6 | 0 | 6.00 |
| 13F | F | 8 | 10 | 8.83 |
| 14M | M | 9 | 1 | 9.08 |
| 15M | M | 9 | 2 | 9.17 |
| 16M | M | 8 | 11 | 8.92 |
| 17M | M | 6 | 2 | 6.17 |
| 18M | M | 7 | 5 | 7.42 |
| 19M | M | 7 | 5 | 7.42 |
| 20M | M | 7 | 11 | 7.92 |
| 21M | M | 6 | 0 | 6.00 |
| 22M | M | 12 | 11 | 12.92 |
| 23M | M | 8 | 4 | 8.33 |
| 24F | F | 11 | 2 | 11.17 |
| 25M | M | 10 | 10 | 10.83 |
| 26F | F | 7 | 4 | 7.33 |
| 27M | M | 7 | 10 | 7.83 |
| 28F | F | 7 | 8 | 7.67 |
| 29M | M | 8 | 2 | 8.17 |
| 30F | F | 6 | 3 | 6.25 |
| 31M | M | 9 | 2 | 9.17 |
| 32M | M | 10 | 9 | 10.75 |
| 33M | M | 5 | 9 | 5.75 |
| 34F | F | 9 | 0 | 9.00 |
| 35M | M | 5 | 11 | 5.92 |
| 36M | M | 6 | 4 | 6.33 |
| 37F | F | 5 | 2 | 5.17 |
Sessions
Each child recorded only one session.
Data Types
Core data types:
| Data type | Description |
|---|---|
| wav | speech waveform |
| ult | raw ultrasound data |
| param | ultrasound parameters |
| txt | prompt text with date/time of utterance recording |
Hardware synchronisation failed for this dataset. We release the parameter File as exported from the AAA software.
Additional data:
| Data type | Description |
|---|---|
| slt_labels | manual annotation from SLT, when available. |
| probe_direction_labels | a label for each File ID indicating whether the probe was in coronal position (cor) or midsagittal (sag) |
SLT Labels are available in Praat's TextGrid format, and probe_direction_labels is a csv file.
File IDs
Individual recordings are indexed according to their recording times. See the prompt text file for recording date/time.
Each file ID also includes a prompt type identifier. See Data for details.
References
[1] Eshky, A., Ribeiro, M. S., Cleland, J., Richmond, K., Roxburgh, Z., Scobbie, J., & Wrench, A. (2018) Ultrasuite: A repository of ultrasound and acoustic data from child speech therapy sessions. Proceedings of INTERSPEECH. Hyderabad, India.