This page tries to document the various ways of getting speech recognition to work under Ubuntu. Feel free to extend it!

What is speech recognition?

Speech recognition under Ubuntu

Open source solutions

There is no fully functional open source solution yet, but you can train one of the following packages to your voice. Moreover, you can help the development of open source speech recognition softwares by recording your voice in the Voxforge project.

HTK and Julius/Julian

This tutorial is based on the extensive HTK+Julius tutorial on the Voxforge site, and highlights only the necessary changes to make it working.

Download and install the programs

• The versions linked from the tutorial are outdated, and HTK won't compile. Please use a recent version. To compile HTK you will need probably the following header packages: libx11-dev, libxtst-dev. Install them using your favourite package manager. Install audacity using your favourite package manager.

Step 2 - Pronunciation Dictionary

• The perl scripts look for the perl interpreter under /usr/local/bin instead of /usr/bin. Change the scripts' first line to

Proprietary solutions

Dragon Naturally Speaking

Dragon Naturally Speaking is probably the most widely used speech recongition software, especially under Windows. Unfortunately the recent version (Dragon Naturally Speaking 9) does not have a Linux version. Previous version are available for Linux as well.

Howto

IBM ViaVoice

It a speech recognition software intended for embedded devices, not for desktops.

Howto

CategoryNeedsExpansion