AudioSpecial Audio/Music

1,300-person voiceprint speech dataset

About this dataset

1. Product Overview

The voice database comprises a total of 602 hours of recordings. The recording language is Chinese, and the recordings were made in China. A total of 1,300 speakers from various accent regions across China were invited; each speaker recorded three times—morning, noon, and evening—on different days, with individual recording sessions spanning 7 to 78 days. The speakers’ ages range from 9 to 65 years old, and recordings were captured using mobile phones, covering commonly used smartphone models. This database can be utilized for research such as speaker recognition. Below is a table showing the time intervals between the three recordings for each of the 1,300 speakers, where t1 denotes the interval between the first and second recordings, and t2 denotes the interval between the second and third recordings.

（Line chart axis scale）
X-axis (recording interval in days, t1/t2): 1, 501, 1001, 1501, 2001, 2501, 3001, 3501, 4001, 4501
Y-axis (number of speakers): 71, 66, 61, 56, 51, 46, 41, 36, 31, 26, 21, 16, 11, 6, 1

2. Product Directory Structure

Directory Structure

Data directory structure

│ spk_info.xlsx（Speaker information）
│ content.txt（Voice text content）
└─SPEECHDATA（Data folder）

│├─wav（Audio folder）

Naming Conventions

\SPEECHDATA\wav\\\ e.g., \SPEECHDATA\wav\0001\0001_F0001.wav

Name · Content · Notes

· 0001–4595 · Speaker ID

· _F/S/T_.wav · Audio file name

· F—First recording, S—Second recording, T—Third recording · Recording sequence

3. Text Design

Considering diverse industrial applications, the voice data was collected in three separate sessions—F, S, and T. The text designs for each session are as follows: F-text: First recording. Each speaker provides 130 sentences consisting of numeric sequences, such as “zero one two three four five six seven eight nine,” with each sentence containing 5–15 digits, averaging around 10 digits. S-text: Second recording. Each speaker provides 160 sentences: sentences 0001–0025 contain control words plus wake-up phrases; sentences 0026–0050 feature free-form text covering sensitive topics like security and medicine; and sentences 0051–0160 consist of broad-topic free-form text. T-text: Third recording. Each speaker provides 130 sentences, all of which are broad-topic free-form text.