Skip to content
Automatic Speaker Recognition algorithms in Python
Branch: master
Clone or download
orchidas Merge pull request #3 from chrisspen/master
Fixed formatting and Python3 errors.
Latest commit a12467e Mar 19, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
test
train Add folder containing sound files to train on - the word 'zero' utter… Apr 3, 2016
.gitignore Fixed formatting and Python3 errors. Mar 18, 2018
ASR_report.pdf
LBG.py Fixed extra tabs in LBG.py Apr 3, 2016
LICENSE
LPC.py Add program to calculate Linear Prediction Coefficients Apr 3, 2016
README.md Fixed formatting and Python3 errors. Mar 18, 2018
__init__.py
mel_coefficients.py
requirements.txt Fixed formatting and Python3 errors. Mar 18, 2018
spectrogram.py
test.py
train.py Fixed formatting and Python3 errors. Mar 18, 2018

README.md

Speaker-Recognition

Automatic Speaker Recognition algorithms in Python

This repository contains Python programs that can be used for Automatic Speaker Recognition. ASR is done by extracting MFCCs and LPCs from each speaker and then forming a speaker-specific codebook of the same by using Vector Quantization (I like to think of it as a fancy name for NN-clustering). After that, the system is trained and tested for 8 different speakers.

Create virtualenv with:

virtualenv -p python3 .env
. .env/bin/activate
pip install -r requirements.txt

To test the algorithm, run test.py. Certain parameters are open to be changed, such as the order of LPC coefficients, the number of Mel filterbanks and the number of centroids in each codebook. Everything is included in the repository, including .wav files for testing and training, hence cloning it and running test.py should work.

A PDF has been included that explains the theory and provides links to relevant websites and projects.

You can’t perform that action at this time.