Home Scene Voice-Controlled Word Speech Database
About this dataset
1. Product Overview
The home‑scene voice control vocabulary speech database was created by inviting 399 participants to record in a quiet environment using high‑fidelity microphones. A total of seven recording positions were used, with each position capturing 80.78 hours of audio, amounting to 565 hours overall. The sampling rate is 48 kHz and the bit depth is 16 bits. This database has been transcribed and annotated by professional speech‑recognition experts and subjected to rigorous quality checks, achieving a text accuracy rate exceeding 95%.
2. Speaker Information
Gender Distribution
The database includes a total of 399 speakers: 209 males and 190 females.
Gender · Male · Female · Total
Proportion · 52% · 48% · 100%
Age Distribution
The speakers’ ages range from 16 to 50 years old, distributed as follows:
· Age Group · Number of Speakers · Proportion
A · 16–25 years · 149 · 37.3%
B · 26–40 years · 185 · 46.4%
C · 41–50 years · 65 · 16.3%
Total · 399 · 100% ·
Accent Distribution
The speakers are evenly distributed across seven regions, as follows:
No. · Region · Number of Speakers
1 · Fujian · 57
2 · Guangdong · 57
3 · Guangxi · 57
4 · Jiangsu · 57
5 · Shanghai · 57
6 · Sichuan · 57
7 · Zhejiang · 57
Total · · 399
3. Recording Environment
Recording Venue
The recordings were conducted in quiet indoor settings across seven cities: Guangdong, Guangxi, Sichuan, Fujian, Jiangsu, Zhejiang, and Shanghai.
Recording Equipment
High‑fidelity microphones and audio recorders were used for recording. All data in this database is stored in mono format at 48 kHz and 16 bits.
Recording Method
Seven recording positions were set up indoors, including one close‑talk position at 0.5 meters and six functional positions. The functional positions were fixed at 1 meter, 3 meters, and 5 meters, with two microphones at each position spaced 15 cm apart, as shown in the figure below:
4. Speech Data Transcription
Transcribers render the spoken content based on the audio they hear, striving to ensure that the written text accurately reflects the original pronunciation. General guidelines include: 1) The transcription must exactly match the spoken words—no extra or missing words, and no errors. 2) Numbers should be written in Chinese characters, such as “一二三,” rather than “123.” 3) English words should be written in lowercase; proper nouns and names should be capitalized with spaces, e.g., “W W W dot Google dot com” or “C E O.” 4) Annotations must fully reflect the actual pronunciation without any omissions.
5. Data File Directory
Directory Structure
The data directory tree is organized as follows:
└─DOC (textual documentation)
││├─wav_list.txt (audio file list)
││├─content.txt (transcription list)
││├─spk-info.xlsx (speaker information table)
└─SPEECHDATA (audio folder)
││├─M0001 (speaker ID)
││├─M0001.txt (transcribed text)
││├─wav (audio folder)
││ ││ M0001P01W0001.wav (audio file)
Naming Conventions
CORPUS/USAGE/SPEAKER_ID/AUDIO_DIR/SPEECH_IDe.g./SPEECHDATA/M0001/wav/M0001P01W0001.wavDirectory NameContent Notes
CORPUS · · Speech Database Number
USAGE · SPEECHDATA · Audio Folder
SPEAKER_ID · M0001 · Speaker ID
AUDIO_DIR · wav · Audio Folder
SPEECH_ID · M0001P01W0001.wav · Audio File
Data Sample Download
Samples
- Audio sample 1