David Huggins-Daines <dhuggins@cs.cmu.edu>
http://www.cs.cmu.edu/~dhuggins
January 7, 2008
This module requires PocketSphinx 0.4.1 or later, available from http://www.sourceforge.net/projects/cmusphinx/. To build and install PocketSphinx, download sphinxbase and pocketsphinx and untar the packages. Now run:
cd sphinxbase-0.3
./configure
make
# now become root, or some user with write access to /usr/local
make install
cd ../pocketsphinx-0.4.1
./configure
make
# now become root, or some user with write access to /usr/local
make install
On GNU/Linux systems, you may need to add /usr/local/lib to the end of your /etc/ld.so.conf file and re-run /sbin/ldconfig as root.
Once you have installed PocketSphinx, you can build and install Speech::Recognizer::SPX. To do this, change into the Speech-Recognizer-SPX directory, and run:
perl Makefile.PL
make
make test # this may fail if you have no working audio device
make install
If any failures are reported by 'make test' that are not audio-related, please report them by e-mail to the address above.
This module has only been tested on GNU/Linux. It is likely to work on free 4.4BSD-derived operating systems as well, but I have not tested this. Status on other Unix-like operating systems supported by PocketSphinx, or Microsoft Windows, is unknown.
Several example programs are included in this distribution:
tty-continuous: Basically the same as the 'pocketsphinx_continuous' demo
included with the PocketSphinx distribution.
turtle-demo: Basically the same as the 'pocketsphinx_demo' program included
with the Sphinx-II distribution.
server-demo: Demo of using a separate process to do recognition.
fe-test: Demo of the feature extraction model.
select-cad: Demo of using select() on audio devices in conjunction
with Audio::SPX::Continuous.