Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/fighting41love/zhvoice
Chinese voice corpus. 中文语音语料,语音更加清晰自然,包含8个开源数据集,3200个说话人,900小时语音,1300万字。
https://github.com/fighting41love/zhvoice
Last synced: about 1 month ago
JSON representation
Chinese voice corpus. 中文语音语料,语音更加清晰自然,包含8个开源数据集,3200个说话人,900小时语音,1300万字。
- Host: GitHub
- URL: https://github.com/fighting41love/zhvoice
- Owner: fighting41love
- Created: 2020-06-13T05:50:16.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2020-06-12T06:46:45.000Z (over 4 years ago)
- Last Synced: 2024-08-02T18:43:26.401Z (5 months ago)
- Homepage:
- Size: 45.9 KB
- Stars: 555
- Watchers: 9
- Forks: 111
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
- StarryDivineSky - fighting41love/zhvoice
README
![zhvoice](zhvoice.png)
# 中文语音语料
[**zhvoice**](https://github.com/KuangDD/zhvoice): Chinese voice corpus
tips: 中文或汉语的语言简称缩写是**zh**。
喜欢请**star**!你就是**superstar**!
## 语料简介
zhvoice语料由8个开源数据集,经过降噪和去除静音处理而成,说话人约3200个,音频约900小时,文本约113万条,共有约1300万字。
zhvoice语料比较原始数据而言,更加清晰和自然,减少了噪声的干扰,减少了因说话人说话不连贯造成的不自然。
zhvoice语料包含文本、语音和说话人3个方面的信息,可适用于多种语音相关的任务。
zhvoice语料由[智浪淘沙](https://github.com/zhilangtaosha)清洗和处理。
## 处理方法
- 用python的工具模块[**aukit**](https://github.com/KuangDD/aukit)处理音频,降噪和去除静音。```
pip install aukitfrom aukit import remove_noise, remove_silence
```- 用python的工具模块[**phkit**](https://github.com/KuangDD/phkit)处理文本,文本正则化和汉字转拼音。
```
pip install phkitfrom phkit import text_to_sequence, pinyin
```## 应用场景
- 用于语音克隆模型,可直接用于githup的语音克隆项目[**zhrtvc**](https://github.com/KuangDD/zhrtvc)。
- 用于语音合成模型,用标贝开源的中文标准女声音频**zhbznsyp**数据集,或者筛选音质较好,和目标声音相似的说话人语音及其文本。
- 用于声码器模型,即由语音特征转为语音信号的模型。用语音数据,可结合[**aukit**](https://github.com/KuangDD/aukit)的音频转频谱。
```
from aukit import linear_spectrogram, mel_spectrogram, world_spectrogram
```- 用于语音编码器模型,即把语音编码到预定维度的向量空间。
- 用于声纹识别模型,用语音和对应的说话人标签。
- 用于语音识别模型,用语音和文本,可以适当加噪声。## 下载路径
百度网盘:
链接: https://pan.baidu.com/s/1uHXE2WIt0kdm_dPSej-TtA
提取码: i5b3
## 文件介绍
- info:各个数据集的源数据信息,包含源数据出处、简介等。
- text:语音语料对应的文本,包含文本、相对路径、说话人、参考拼音等信息。
- sample:样本语音,每个说话人一个音频。
- metadata:语料元数据,一行对应一个音频文件,每行的格式`音频相对路径\t汉字文本\n`。
- zh*:zh开头的是语料文件,目录结构:根目录下包含metadata.csv和语音文件目录。一个说话人对应一个子目录,音频是mp3格式。metadata.csv的数据结构和metadata的一样,记录当前数据集的信息。## 统计信息
- character_W: 字符个数,单位:万字。包括汉字、英文字母和标点符号。
- duration_H: 语音时长,单位:小时。
- n_audio_per_speaker:每个说话人的音频数量。
- n_minute_per_speaker:平均每个说话人的音频总时长,单位:分钟。
- n_speaker:说话人个数。
- sentence_W:文本数目,单位:万条。
- size_MB:音频占用存储空间,单位:MB。注意:
1. **total**是全部数据集合集的结果。
2. 音频的采样率是16k。```
{
"total": {
"character_W": 1287.0836999999997,
"duration_H": 889.7492555555556,
"n_audio_per_speaker": 348.30255463219453,
"n_character_per_sentence": 11.37366465335554,
"n_minute_per_speaker": 16.431195855134913,
"n_second_per_audio": 2.8305039345725436,
"n_speaker": 3249,
"sentence_W": 113.1635,
"size_MB": 9164.134941101074
},
"zhaidatatang": {
"character_W": 233.5123,
"duration_H": 145.47232,
"n_audio_per_speaker": 395.4416666666667,
"n_character_per_sentence": 9.841835078920194,
"n_minute_per_speaker": 14.547232000000001,
"n_second_per_audio": 2.2072381177164773,
"n_speaker": 600,
"sentence_W": 23.7265,
"size_MB": 1498.3187255859375
},
"zhaishell": {
"character_W": 204.0219,
"duration_H": 142.5542,
"n_audio_per_speaker": 354.0,
"n_character_per_sentence": 14.40832627118644,
"n_minute_per_speaker": 21.38313,
"n_second_per_audio": 3.6242593220338986,
"n_speaker": 400,
"sentence_W": 14.16,
"size_MB": 1468.2630157470703
},
"zhbznsyp": {
"character_W": 18.3708,
"duration_H": 10.544652222222222,
"n_audio_per_speaker": 10000.0,
"n_character_per_sentence": 18.3708,
"n_minute_per_speaker": 632.6791333333333,
"n_second_per_audio": 3.7960748,
"n_speaker": 1,
"sentence_W": 1.0,
"size_MB": 108.60657119750977
},
"zhmagicdata": {
"character_W": 567.2561,
"duration_H": 406.01905,
"n_audio_per_speaker": 563.8938053097345,
"n_character_per_sentence": 9.891471367789634,
"n_minute_per_speaker": 23.953926253687317,
"n_second_per_audio": 2.548769930947897,
"n_speaker": 1017,
"sentence_W": 57.348,
"size_MB": 4181.867351531982
},
"zhprimewords": {
"character_W": 105.2203,
"duration_H": 81.30301,
"n_audio_per_speaker": 171.96621621621622,
"n_character_per_sentence": 20.67115241051432,
"n_minute_per_speaker": 16.480339864864863,
"n_second_per_audio": 5.750085183293388,
"n_speaker": 296,
"sentence_W": 5.0902,
"size_MB": 837.3951988220215
},
"zhspeechocean": {
"character_W": 3.1078,
"duration_H": 1.8908433333333334,
"n_audio_per_speaker": 120.0,
"n_character_per_sentence": 12.949166666666668,
"n_minute_per_speaker": 5.67253,
"n_second_per_audio": 2.836265,
"n_speaker": 20,
"sentence_W": 0.24,
"size_MB": 19.475086212158203
},
"zhstcmds": {
"character_W": 111.9317,
"duration_H": 74.53628,
"n_audio_per_speaker": 120.0,
"n_character_per_sentence": 10.909522417153998,
"n_minute_per_speaker": 5.230616140350877,
"n_second_per_audio": 2.6153080701754385,
"n_speaker": 855,
"sentence_W": 10.26,
"size_MB": 767.7000274658203
},
"zhthchs30": {
"character_W": 43.6628,
"duration_H": 27.4289,
"n_audio_per_speaker": 223.13333333333333,
"n_character_per_sentence": 32.61338512100388,
"n_minute_per_speaker": 27.4289,
"n_second_per_audio": 7.375563190917239,
"n_speaker": 60,
"sentence_W": 1.3388,
"size_MB": 282.5089645385742
}
}
```