https://github.com/pmdartus/htk-digit

Experiment HTK toolkit to for voice recognition
https://github.com/pmdartus/htk-digit

Last synced: 8 days ago
JSON representation

Experiment HTK toolkit to for voice recognition

• Host: GitHub
• URL: https://github.com/pmdartus/htk-digit
• Owner: pmdartus
• Created: 2014-12-04T06:24:48.000Z (almost 10 years ago)
• Default Branch: master
• Last Pushed: 2015-01-05T04:56:44.000Z (almost 10 years ago)
• Last Synced: 2024-10-18T19:01:18.118Z (28 days ago)
• Language: Shell
• Homepage:
• Size: 34.7 MB
• Stars: 5
• Watchers: 2
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Voice Digit Recognition

Use the HTK toolkit to train a model to recognize the spoken numbers.

## Install

Installation of HTK toolkit differs a lot on every platform. The project comes with a `Vagrantfile` in order to ease the project's setup. Vagrant will provision a virtual machine containing Ubuntu 32bit and also install HTK. Refers to [Vagrant's documentation](https://www.vagrantup.com/) for more details.

Because also of the distribution policy of the HTK, you will have to register on the [website](http://htk.eng.cam.ac.uk/) and download the package. Add the compressed package (`HTK-3.4.1.tar.gz`) in the `provision folder`.

You are now ready to run Vagrant.

```

vagrant up

vagrant ssh

cd /vagrant

```

## Usage

You can create a model and train it using the command `./start-htk.sh`. The final model will be stored in the folder `Models/hmm15`

The command will also output accuracy of the trainned model against the `dev` dataset.

You can specify the output file for logging the script by passing the name of the file: `./start-htk.sh cmd.log`. If not file is passed the command outputs will be redirected into `/dev/null`.

## Results

Read the [report](REPORT.md)! ;)