PyCon 2010 talk on Pocketsphinx

This is a talk by David Huggins-Daines on Pocketsphinx and Python on PyCon 2010 in Atlanta

It contains quick and nice introduction in Pocketsphinx API with all major issues covered. It will let you write your own speech recognition in 5 minutes.

Category:

Science & Technology

Tags:

• pocketsphinx
• cmusphinx
• python
• speech recognition

[alex1212a]

Convert mp3 to WAV-------------- on Ubuntu-------------

apt-get -y install sox libsox-fmt-mp3

# convert mp3 to WAV (16kHz)

# sox -r 16000 -2 -s -t mp3 mp3filename wavfilename

sox -r 16000 -2 -s -t mp3 rec0505-004232.mp3 goforward.wav

[alex1212a]

I used Google Search:

+"asr.py" +sphinx

[alex1212a]

Installation Procedure for Ubuntu 11-04 (worked the same on 10-10 Ubuntu, just the same)

# install packages (Acoustics Model, Language Model, and a decoder)

sudo apt-get install python-pocketsphinx pocketsphinx-lm-wsj pocketsphinx-hmm-wsj1 ipython

# sudo apt-get -y install vim ssh

Created one Python file, and one audio file (16KHz); intelligibility (specifically consonants) is considerably improved with the band with afforded by a 16kHz sample rate.The path2audio is in asr.py