ChildMandarin
Comprehensive Mandarin speech dataset for children aged 3-5, with 397 speakers and 41.25 hours of conversational speech
Duration
41.25 hours
Languages
1
Sample Rate
16 kHz
Published
2024-09
Description
1Total duration of 41.25 hours, containing 40,913 utterances with an average length of 3.52 seconds
2397 child speakers aged 3 to 5, with balanced gender distribution
3Speakers from 22 out of 34 provincial-level administrative regions in China
4Accent levels classified as Heavy (H), Medium (M), and Light (L); approximately 95.97% of speakers have light accents
5Recorded on smartphones (216 Android and 181 iPhone devices) in quiet indoor environments
6Audio format: WAV PCM, 16 kHz sample rate, 16-bit precision
7Character-level manual transcription by professional annotators
8Data collected in conversational natural interaction scenarios with parents present for emotional support
9Split into training (317 speakers, 33.35 hours), validation (39 speakers, 3.78 hours), and test (41 speakers, 4.12 hours) sets with no speaker overlap
Language Details
| Language | Duration |
|---|---|
| Mandarin Chinese | 41.25 hours |
Publisher