简体   繁体   English

从音频中提取 ivector

[英]ivector extraction from audio

Good afternoon,下午好,

I am looking at extracting ivectors from audio clips.我正在研究从音频剪辑中提取 ivector。 Is python-bob the best library to use? python-bob 是最好用的库吗? Having difficulty finding any useful documentation so would appreciate any points in the right direction.很难找到任何有用的文档,因此将不胜感激任何正确方向的观点。

thanks谢谢

Since you are asking a very broad question, I can only give you a broad answer.既然你问的问题很笼统,我只能给你一个笼统的回答。

To extract ivectors, you need to train the background model first.要提取 ivector,需要先训练背景 model。 bob.bio.spear is the main library that you can use to do this. bob.bio.spear是您可以用来执行此操作的主库。 bob.db.voxforge can be used as a toy database example to get you started. bob.db.voxforge可以用作玩具数据库示例来帮助您入门。 bob.bio.spear was removed from Bob 9 release (it's being revived for the next release of Bob) so you have to install bob 8 and read bob 8 documentation bob.bio.spear 已从 Bob 9 版本中删除(它正在为 Bob 的下一个版本恢复),因此您必须安装 bob 8 并阅读bob 8 文档

So here is how to install bob 8 relevant packages:所以这里是如何安装 bob 8 相关包:

$ conda create --name bob-8 --override-channels \
  -c https://www.idiap.ch/software/bob/conda -c defaults \
  -c https://www.idiap.ch/software/bob/conda/label/archive \
python=3 bob=8 bob.bio.spear bob.db.voxforge gridtk bob.bio.gmm
$ conda activate bob-8
$ conda config --env --add channels defaults
$ conda config --env --add channels https://www.idiap.ch/software/bob/conda/label/archive
$ conda config --env --add channels https://www.idiap.ch/software/bob/conda
$ # conda install more packages if you need them. like:
$ # conda install bob.db.iris

Here are some useful links that can help you get started:以下是一些可以帮助您入门的有用链接:

If you want to look at the code at Gitlab, you have to browse the code in the tag which is associated with Bob 8 release: https://gitlab.idiap.ch/bob/bob/-/blob/v8.0.0/requirements.txt如果您想查看 Gitlab 的代码,您必须浏览与 Bob 8 版本相关的标签中的代码: https://gitlab.idiap.ch/bob/bob/-/blob/v8.0.0/要求.txt

For example, to browse bob.bio.base code, look at https://gitlab.idiap.ch/bob/bob.bio.base/-/tree/v4.1.1例如,要浏览 bob.bio.base 代码,请查看https://gitlab.idiap.ch/bob/bob.bio.base/-/tree/v4.1.1

For a minimal example, to train and test and i-vector system on the voxforge database, run:举个简单的例子,要在 voxforge 数据库上训练和测试 i-vector 系统,运行:

$ download_and_untar_voxforge.py --address ~/databases/voxforge
$ echo "[YOUR_VOXFORGE_DIRECTORY]=$HOME/databases/voxforge" > ~/.bob_bio_databases.txt
$ verify.py -vvv -d voxforge -p energy-2gauss -e mfcc-60 -a ivec-plda-voxforge -s ivec-plda --groups {dev,eval} --parallel 4

then, if you want to test on your own data, you can create a filelist database interface following https://www.idiap.ch/software/bob/docs/bob/docs/v8.0.0/bob/bob.bio.base/doc/filelist-guide.html然后,如果您想对自己的数据进行测试,您可以按照https://www.idiap.ch/software/bob/docs/bob/docs/v8.0.0/bob/bob.bio 创建一个文件列表数据库接口。基础/doc/filelist-guide.html

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM