Skip to content

speaker diarization in phone recording/电话录音中的说话人分离

Notifications You must be signed in to change notification settings

IvanEvan/speaker-diarization

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

1. Overview

 This repo was created to separate two speakers from a telephone recording.
 If your telephone recording has more than two speakers, I can't guarantee that my method will work.
 In addition to this one, to get good result, please try to make sure that different speakers have the same length of speech.


2. Implement

 1. Split a wave to audio clips by remove mute
 2. Count all clips' id-vector use pre-trained speaker recognition model
 3. Use K-means to cluster all clips' id-vector when K=2


3. Result

image


4. Appendix

 1. The pre-trained speaker recognition model from WeidiXie's repo VGG-Speaker-Recognition. Thanks for the open source!
 2. Because my method looks like a non-supervised method, so you can try supervised method even end2end. You can get more information about speaker diarization from Here

About

speaker diarization in phone recording/电话录音中的说话人分离

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages