I've summarized here the great article about transcription process using the CMUSphinx by Andrew Brobinson[1]. Enjoy it!
On your linux terminal (Tested in Ubuntu 12.10 and Debian 6.0), type the following commands
1 - sudo apt-get install sphinx*
2 - sudo apt-get install libcdio-paranoia-dev
3 - sudo apt-get install python-all-dev
4 - wget http://downloads.sourceforge.net/project/audiotools/audiotools/2.16/audiotools.2.16.tar.gz
5 - tar -xvf audiotools.2.16.tar.gz
6 - cd audiotools.2.16
7 - sudo make install
8 - sudo python setup.py build
9 - sudo python setup.py install
Before to start, you need to download an acoustic model,a language model and a dictionary for your language. At CMUSphinx site[2], you have available ( ready-to-use) models for English, Russian, Spanish, German and Dutch.
Finally, edit the speech-to-text.py file with your parameters (folders and audio file) and run
python speech-to-text.py
[1] - http://andrewbrobinson.com/2012/02/29/speech-recognition-using-sphinx-dont-try-this-at-home/