John Wojnaroski wrote:
----- Original Message -----
From: "David Luff" <[EMAIL PROTECTED]>
This all sounds very exciting, especially the encouraging results from the
voice recognition stage, and the fact that Jon thinks that Festival 2 is
sounding pretty good. Could you send me the code you've got so far for
sending strings across to FG?
Actually, there are three parts --- the ASR that converts speech to text
(that runs on my laptop) and sends the text string over the LAN to the AI
app that analyzes and generates a response and sends the text over to the
festival engine (TTS) for conversion back to audio. I'll send along more
details via private mail and attach the tar files.
What approach do you currently use to simulator the 2nd and
most interesting ;-) part ?
on the decoding of the speech to text-strings only so far, or have you
actually started on logically decoding the text strings for ATC-AI? This
is the part I'm currently in deep thought about.
Working on the speech to text and text to speech ends, the middle ATC/AI is
the really tough part.
yep, since the day you posted your idea I've also been playing around
with festival locally - and to be honest, I just used nested switch
statements so far for the textual recognition part - but hey,
that's also a (simple) "state machine" ;-) , simply because it's really
getting pretty abstract without using some kind of more advanced AI
mechanism/library, I've just looked into libs that support basic FSM's -
but in most cases a finite state machine would be also a pretty static
solution :-/
So, you'd have to create at state machine definition file, and have
a pre-processor then compile this into C++ source code - hence this
would be compile-time static, but one would really need to resemble a
basic state machine manually using classes that access XML files,
in order to keep things dynamic, and in order not to have to
recompile the "AI"-part for each new phraseology item.
I am just about to check boost.org for templates that might be
suitable for the FSM modelling part. If anybody knows about
C++ libs/templates for dealing with state machines, it might
certainly be interesting to look into that option.
Otherwise, there are numerous simple editors for the creation of
finite state models, that one could use to create a basic framework -
but taking into account that this is not only about recognition of
a message, but also about properly dealing with the parameters it's
getting really abstract - basically one would nett pretty much
a pendant of what "SABLE" is doing for festival, for the
recognition engine, too.
Have you ever looked at www.vatsim.org or www.ivao.org ?
I have, and I know people who are 'active virtual controllers'
there ...
A different approach using real "actors" to create a virtual
world.
...and they'd tell you it's not only a different approach
it's also a different MOTIVATION - they don't do it because
they are interested in FLIGHT SIMS, but rather they are ATC
enthusiasts - so it's not about 'acting', but rather they
seem to enjoy virtual controlling as much as others enjoy
virtual flying ...
But, alas, it's all MS based...
well, actually not really - at least the protocols are meanwhile "open"
and there are several attempts ongoing to establish some kind of
cross-platform tool, I think one can even find several projects like
that on sourceforge.
The only problem is not so much the ATC client/server part,
which is so far mainly about a simple VoIP tool, but rather the real
issue is multiplayer integration *and* simulator data fetching -
of course Micro$oft is leading in the multiplayer area - and
fsuipc does the rest: there are thousands of people who "fly online"
using Microsoft's sim each day - compared to those other minorities
that use different simulators, and even *different protocols*.
So even if a flightsimulator like FG had already serious multi-player
capabality it would certainly not be that simple to make FG interact
with all those version of M$ FS200x.
If it was only for the client tools that are used by the
"pilots" & "controllers", this whole thing would be a lot easier,
but since there needs to be basic multiplayer functionality, as well
as data extraction from the simulator (for the virtual radar screens),
it's indeed somewhat restricted right now to software like Micro$oft FS.
Flight simulators, as well as ATC simulators have been around for
quite a while, the latter of course with not such significant success,
but it's getting more popular each day because the "fun factor" is
increasing, simply because of technologies like TTS/ARS - meanwhile
ATC'ing is not merely about using the keyboard & mouse and doing
abstract thinking, anymore - , but rather you can now use products like
Aerosoft's ATC simulator, that make use of TTS/ARS in order to make the
whole thing more playable.
So, they are now trying to combine both worlds - which of course makes
sense, but is not really THAT trivial.
----------
Boris
_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel
2f585eeea02e2c79d7b1d8c4963bae2d
