Skip to content

Deep learning approach to identifying languages from short audio samples. Method for obtaining and preprocessing a dataset included.

License

Notifications You must be signed in to change notification settings

ibro45/Language-Identification-Speech

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Language Identification From Speech

The repository contains the code that allows you to obtain a dataset for desired languages and train a neural network that's able to identify the language spoken in a recording as short as 3 seconds.

This code follows my bachelor's thesis, "Recognition of Natural Language Type Using Tensorflow", where you can find the detailed description of the project along with the results achieved.

I'd also like to acknowledge that this project was based on the project that this repository has forked from and their paper, "Language Identification Using Deep Convolutional Recurrent Neural Networks".

Repository Structure

The repository is separated into four categories, with each containing its own ReadMe.

Folder Description
data/ Contains the scripts for downloading and preprocessing the dataset.
tensorflow/ Neural-network-related scripts. Creation, training, evaluation etc.
record_predict/ Scripts that allow the user to record his/her speech and test it with the trained model.
paper/ Contains the author's paper titled "Recognition of Natural Language Type Using Tensorflow".

Requirements

To install all the required packages, run

pip install -r requirements.txt

Please note that the requirements.txt specifies the CPU-based tensorflow. If you'd like to train the neural network using GPU(s), check tensorflow-gpu. The installation of the GPU version can be tricky, but using conda and this tutorial may make it a bit easier. Also, make sure that you have the version of Nvidia drivers compatible with the current tensorflow-gpu version.

Furthermore, you should have the following utilities installed in your system:

  • FFmpeg
  • SoX
  • youtube-dl

If you're running Ubuntu, you can install them by running:

sudo apt install ffmpeg sox youtube-dl

About

Deep learning approach to identifying languages from short audio samples. Method for obtaining and preprocessing a dataset included.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%