ROS bindings to pocketsphinx, using gstreamer
Clone or download
Pull request Compare This branch is 12 commits ahead of mikeferguson:indigo-devel.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
demo
language_models
launch
nodes
CHANGELOG.rst
CMakeLists.txt
README.md
package.xml

README.md

A simple ROS wrapper for using Pocketsphinx (via gstreamer) with ROS. See docs here http://wiki.ros.org/pocketsphinx

If installing from source you will need to install the following:

sudo apt-get install gstreamer0.10-pocketsphinx pocketsphinx-lm-en-hub4 pocketsphinx-utils pocketsphinx-hmm-en-hub4wsj ros-indigo-audio-common

Subscribing to ROS audio messages:

To subscribe to ROS audio messages, pass the audio_msg_topic parameter to the node (normally, this is /audio). The recognizer will subscribe to this topic and use the AudioData messages as input to pocketsphinx.

Requires the audio_common package.

Creating a new language model:

Creating a new language is straightforward once you have a corpus of possible sentences your system should handle. To generate the language model and dictionary:

  1. Create <filename>.corpus with one sentence per line.
  2. Upload the corpus file here: http://www.speech.cs.cmu.edu/tools/lmtool-new.html
  3. Download the .dic and .lm file, and rename them appropriately (<filename>.dic).
  4. Change the lm and dict parameters for the recognizer, for example in the launch file.