https://github.com/RicherMans/PLDA

An LDA/PLDA estimator using KALDI in python for speaker verification tasks
https://github.com/RicherMans/PLDA

kaldi plda speaker-verification

Last synced: 2 months ago
JSON representation

An LDA/PLDA estimator using KALDI in python for speaker verification tasks

• Host: GitHub
• URL: https://github.com/RicherMans/PLDA
• Owner: RicherMans
• Created: 2015-10-04T08:52:43.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2017-04-15T03:13:21.000Z (about 7 years ago)
• Last Synced: 2024-01-28T12:39:53.349Z (5 months ago)
• Topics: kaldi, plda, speaker-verification
• Language: Python
• Size: 149 KB
• Stars: 98
• Watchers: 10
• Forks: 24
• Open Issues: 2
• Metadata Files:
  • Readme: README

Lists

• awesome-speaker-recognition-verification - PLDA/LDA from enrollment using Kaldi - PLDA scoring (Table of Contents / Tools)
• my-awesome-stars - RicherMans/PLDA - An LDA/PLDA estimator using KALDI in python for speaker verification tasks (Python)

README

        
=== A simple library to run LDA ===

With this library LDA can be run. The library provides overall just two simple methods to estimate

and predict an LDA classifier.

=== Installation ===

Make sure you have the variable KALDI_ROOT set in your path.

Otherwise:

export KALDI_ROOT=/YOUTPATH/TO/KALDI/ROOT

then just build the libraries:

mkdir build

cd build

cmake ../

make

The installation process compiles a shared library and drops it into the local users python path.

To use it, please just write any python script:

import liblda

lda = liblda.LDA()

#Estimate an LDA model

lda.fit(spktoutts,targetdim)

transformedutt = lda.fit(pathtoutt)

Make sure that spktoutts is a dict where the keys represent the speaker model and the values

represent all the utterances belonging to that speaker. Therefore it is a dict with strings

as keys and a list as values.