Hindi speech data

1. Product Overview

The speech database comprises a total of 568 hours of recordings, in Hindi, recorded in India. The transcribed content covers ten domains, including education, technology, entertainment, health, and more. A total of 580 speakers from various regions across India participated in the recording process, which was conducted in a quiet indoor environment using mobile phones (16 kHz, 16-bit). After transcription and annotation by professional linguists, and following rigorous quality assurance procedures, the database achieves a text accuracy rate exceeding 95%.

2. Recording Corpus

Corpus Construction

Corpus Content

A statistical breakdown of the recording domains is provided below.

Serial No. · Domain

1 · Education

2 · Entertainment

3 · Technology

4 · Health

5 · Lifestyle & Fashion

6 · News

7 · Social Chat

8 · Business & Finance

9 · Sports

10 · Fiction

Corpus Processing

• De-identification: Removal of politically sensitive, personally identifiable, or sexually explicit/violent content.
• Removal of symbols such as <, >, [, ], ~, /, \, =, etc.
• Exclusion of materials containing languages other than Chinese and English.
• Standardization of formatting.

Structural Design of Recording Texts

To ensure comprehensive coverage and phonetic balance, each recording in this database consists of 420 sentences, randomly sampled from the corpus, with the following structure:

Serial No. · Domain · Sentences per Speaker

1 · Education · 40

2 · Entertainment · 30

3 · Technology · 35

4 · Health · 20

5 · Lifestyle & Fashion · 50

6 · News · 80

7 · Social Chat · 120

8 · Business & Finance · 20

9 · Sports · 20

10 · Fiction · 5

Total · 420 ·

3. Speaker Information

Basic Information Records

Speaker information includes task ID, gender, and age range.

Task ID · Gender · Age Range

H0001 · Male · A

Task ID: Each speaker receives one unique task ID, corresponding to a single recording assignment. Each speaker may participate in only one recording session. Age ranges are defined as follows: A (under 18), B (18–25), C (26–40), D (40+).

Speaker Demographic Characteristics

Gender Ratio

Gender · Male · Female · Total

Ratio · 53% · 55% · 100%

Age Distribution

Age Group · Proportion

A · Under 18 · 5%

B · 18–25 · 68%

C · 26–40 · 19%

D · Over 41 · 9%

Total · 100% ·

4. Data Recording Environment

Recording Environment

The recording took place indoors in a quiet setting, free from noticeable background voices or other noise, with no reverberation. Speakers read aloud the prepared texts at their normal speaking pace.

Recording Equipment

Recording equipment included iOS and Android smartphones.

Recording Method

Speakers maintained a distance of approximately 20 cm from the phone, speaking at a normal volume and pace while reading the designated texts.

5. Product Directory Structure

Directory Structure

The data directory structure is organized as follows:

└─DOC（Textual Description Files）
││├─wav_list.txt（Audio File List）
││├─content.txt（Speech Text Content）
││├─spk_info.xlsx（Speaker Information）
│
└─SPEECHDATA（Data Folder）
│├─H0005（Speaker Folder）
││ ││ H0005W0001.wav（Audio File）
││ ││ H0005W0001.txt（Text File）

Naming Conventions

CORPUS/USAGE/SPEAKER_NUM/SPEECH_IDe.g./SPEECHDATA/H0005/H0005W0001.wav

Directory Name · Content · Notes

CORPUS · · Speech Database Identifier

USAGE · SPEECHDATA · Folder Name

SPEAKER_NUM · H0005 · Speaker Folder Name

SPEECH_ID · H0005W0001.wav · WAV File

SENTENCE_ID · H0005W0001.txt · TXT File

Data Sample Download

Click to download the sample dataset (data_sample.zip)