if anyone wants the files (about 200k) give a holler
you can run the whole mess on a single machine along with FG. The hit to the
frame rate is TBD.
----- Original Message -----
From: "John Wojnaroski" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Tuesday, September 21, 2004 9:44 AM
Subject: Voice stuff
> Hi David,
> Attached are two files:
> comm_747 is a really bad hack for the CMU Sphinx ASR engine to create a
> string and sent it out on the network to the "atc_server"
> voice.tgz is more hacking to send a text string to the festival server..
> lines 73 to 76 is my highly sophisticated AI /ATC controller ;-)
> it untars as ..../Voice
> You can run festival as a server "../bin/festival --server" on the same
> machine you run "atc_net_demo" and the audio will be produced on that
> machine or you can uncomment lines 295-309 and also send a .wav file back
> the client machine.
> A quick recap:
> Machine #1 runs sphinx2 which receives the audio input from the user,
> converts it to text and sends it over to Machine #2 which parses the text
> string into tokens and does it's AI/ATC stuff, formulates a text response
> and passes that to the festival server (in this case on 127.0.0.1) which
> creates the audio file and outputs it to the local soundcard. If you run
> festival server on a seperate network machine or back on #1 you need to
> create a short .scm script to add the user to the festival access list as
> You need to upload the ASR sphinx2 stuff and TTS festival stuff from CMU
> wherever. If you need some help or tips in that area give a holler...
> http://linux-sound.org/speech.html is a list of speech related websites
> might find useful. In particular--
> the Festival set, XVoice, Sphinx, and MBROLA. I'm using the
> You'll find http://www.speech.cs.cmu.edu/tools/lmtool.html quite useful
> creating a LM and phonelist for the ASR program. With the smaller
> dictionaries voice recognition is very good but if you mumble a lot (like
> do) the XVoice folks have a wiki page with instructions on how to add a
> voice trainer and improve/tailor the AM for individual speech patterns.
> Setting up the programs takes a little work. You can use the AM provided
> with sphinx2, but you'll get much better results if you upload and install
> the hub4-2000-11-17-1 model. And you will need to create a LM that
> ATC phrases and words
> Good luck, again you've got my email, don't hesitate if you need help or
> have questions.
> John W.
Flightgear-devel mailing list