Voice recognition inside docker container
Python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
wav
Dockerfile
README.md
audio.conf
cmu07a.dic
cmusphinx-en-us-5.2.tar.gz
okdocker.py
pocketsphinx-0.8.tar.gz
sphinxbase-0.8.tar.gz

README.md

How to train your Docker (using voice recognition)

This is a work in progress but just wanted to get this code pushed incase someone else finds it useful.

Before you start there is one file that is greater than 100mb and I couldn't push it to github :

http://sourceforge.net/projects/cmusphinx/files/Acoustic%20and%20Language%20Models/US%20English%20Generic%20Language%20Model/cmusphinx-5.0-en-us.lm.gz/download

If you want to build the image locally you will need to download that file and gunzip it. You should end up with this extracted file : cmusphinx-5.0-en-us.lm

The Build and Usage commands are in the Dockerfile.

The blog post associated with this can be found here :

http://thshaw.blogspot.com/2016/01/how-to-train-your-docker-using-voice.html

The docker image can be pulled from : thshaw/okdocker

I'll be updating this code regularly as it is only partially working. More details in the blog.

Here is what currently works :

Recording 3 second clip
Playback clip
Decoding clip to text

This was built and tested on Ubuntu 15.10 and you may need to tweak the audio on the host to get this to work.

This is still a work in progress and the accuracy is very low. If you want to train Sphinx to use your own voice then then instructions can be found here : http://cmusphinx.sourceforge.net/wiki/tutorialadapt

The next version of the code will use pocketsphinx_continuous and gstreamer : http://cmusphinx.sourceforge.net/wiki/gstreamer

Rgds, @tomwillfixit