QtSpeechRecognition

Qt wrapper around pocketsphinx with basic functionality focused on keywords. Currently only tested and working on Unix systems. Windows is planned for the future

Requirements

This requires CMUSphinx to be compiled/installed on your system. This takes inspiration from the abandoned QtSpeech project to include Speech-To-Text functionality (currently it only has Text-To-Speech functionality).

While you can just compile and install CMUSphinx without any modications, I highly recommend applying the pocketsphinx_qt.patch located in the main directory. It adds the functionality to use keywords from memory rather than requiring a file to read from.

You can follow the installation from the CMUSphinx Tutorial to install PocketSphinx on your system. Or follow the instructions in the next section.

Basic CMUSphinx Installation Tutorial

Installing SphinxBase

1. clone https://github.com/cmusphinx/sphinxbase
2. go into the sphinxbase folder and run ./autogen.sh
3. ./configure then make and make install

Installing PocketSphinx

1. clone https://github.com/cmusphinx/pocketsphinx
2. copy the pocketsphinx_qt.patch located in the QtSpeechRecognition folder, into the pocketsphinx folder
3. go into the pocketsphinx folder and run patch -p1 < pocketsphinx_qt.patch to patch the files.
4. run ./autogen.sh
5. ./configure then make and make install

Post Installation

Setup Library Paths

CMUSphinx installs the libraries by default in the /usr/local/lib/ directory. Most systems do not load libraries from this directory. You can either move them to the normal library location, or do something like:

1. Editing /etc/ld.so.conf to contain /usr/local/lib directory, and then sudo ldconfig to update the config
2. Or by exporting:

export LD_LIBRARY_PATH=/usr/local/lib
export PKG_CONFIG_PATH=/usr/local/lib/pkgconfig

See the CMUSphinx Tutorial if you have more questions regarding setting up the libraries.

CMUSphinx Models

The pocketsphinx folder contains a model folder, containing the models needed for speech recognition. Drop them in the same directory as the DLLTestLoader application (I should've renamed it to Demo Instead). It should load the model just fine.

If the model doesn't load, check the path to laod the model when initializing the speech recognition library.

Including CMUSphinx with QtSpeechRecognition

Simple solution for including CMUSphinx files for compile is to look in the QtSpeechRecognition folder. There's two empty folders here with the names pocketsphinx and sphinxbase. Drop your respective pocketsphinx and sphinxbase cloned repos in the same place and the default links will work just fine