xPL Automatic Speech Recognition with Julius
I’ve wrote a very small Perl module for interfacing my home automation Speech Recognition Engine based on Julius with the xPL Network.
The goal of this xpl-perl module, is to broadcast recognised speech.
I’ve described the ASR xPL schema on the project forum : ASR.BASIC Schema proposal
You need a working Julius installation running in monitor mode (listening on the network) and an xpl-perl setup.
Here is an sample command output :
$ julius -input file -C julian.jconf
...
Stat: server-client: socket ready as server
///////////////////////////////
/// Module mode ready
/// waiting client at 10500
///////////////////////////////
In others windows, you must run xpl-asr-julius and xpl-logger.
When you speak or send a wav file to Julius, the reconised text is broadcasted on the network :
### read waveform input
enter filename-> test.wav
Stat: adin_file: input speechfile: test.wav
STAT: 180003 samples (3.75 sec.)
STAT: ### speech analysis (waveform -> MFCC)
STAT: 00 _default: 17 generated, 17 pushed, 6 nodes popped in 1123
xpl-logger output :
10.0.0.242:53922 [xpl-trig/asr.basic: bnz-julius.nux -> * - <s> éteindre lumière chambre </s>]
Installation
The setup is very simple on Debian GNU/Linux. Just fallow thoses instructions :
$ wget /wp-content/contrib/xpl-asr-julius/xPL-ASR-Julius-0.01.tar.gz
$ tar xzf xPL-ASR-Julius-0.01.tar.gz
$ dh-make-perl xPL-ASR-Julius-0.01
$ cd xPL-ASR-Julius-0.01
$ dpkg-buildpackage -b
# dpkg -i ../libxpl-asr-julius-perl*.deb
That’s all !
