voicemap

This repository contains code to build deep learning models to identify different speakers based on audio samples containg their voice.

The eventual aim is for this repository to become a pip-installable python package for quickly and easily performing speaker identification related tasks.

This tensorflow/Keras/python2.7 branch is discontinued. Work is continuing on the pytorch-python-3.6 branch which will become the master branch.

Instructions

Requirements

Make a new virtualenv and install requirements from requirements.txt with the following command.

pip install -r requirements.txt

This project was written in Python 2.7.12 so I cannot guarantee it works on any other version.

Data

Get training data here: http://www.openslr.org/12

• train-clean-100.tar.gz
• train-clean-360.tar.gz
• dev-clean.tar.gz

Place the unzipped training data into the data/ folder so the file structure is as follows:

data/
    LibriSpeech/
        dev-clean/
        train-clean-100/
        train-clean-360/
        SPEAKERS.TXT

Please use the SPEAKERS.TXT supplied in the repo as I've made a few corrections to the one found at openslr.org.

Run tests

This requires the LibriSpeech data.

python -m unittest tests.tests

Contents

voicemap

This package contains re-usable code for defining network architectures, interacting with datasets and many utility functions.

experiments

This package contains experiments in the form of python scripts.

notebooks

This folder contains Jupyter notebooks used for interactive visualisation and analysis.