Fwd: Speech Recognition application.

Phill Whiteside phillw at thesii.org
Sun Dec 4 01:56:49 UTC 2011

Hi guys & gals...

I've looked after that little phoenix, time to awaken it. I know it and SII
are "upstream" but... any volunteers to assist.

Hi folks,
> As some of you know, SpeechControl has been through a bad patch. However,
> it's a requirement of the SII as a foundation to have SpeechControl working
> as a team once again. Can those who wish to drive SpeechControl up from its
> ruins please make yourselves known by replying to this message with a
> confirmation along with a listing of skills that you have? This will pull
> the few developers we have of the Wintermute project. It is thought that
> after a period of getting this up and running, this project can be
> maintained by one person. Guess what I'm looking for :)
> For the SII team, they do have a decent server area the 16G at
> http://www.kimsufi.co.uk/ as Wintermute's AI and requires a lot of
> computational power in order to speed up its growing process. They asked
> for some horses under the bonnet; I can jump to the  24G if it needs.
> Yes, we are upstream; meaning we reach each Linux distribution out there
> (or at least try). But, Speech Control will be restarted I have missed the
> step getting even get Freenode authorisation, which the previous users did
> not as we have #sii authority. Having had a chat, the area will be
> #sii-speechcontrol.
> Regards,
> Phill

---------- Forwarded message ----------
From: Dante Ashton <dante.ashton at thesii.org>
Date: Sat, Dec 3, 2011 at 4:44 PM
Subject: Speech Recognition application.
To: Jacky Alcine <jacky.alcine at thesii.org>, Phill Whiteside <
phillw at thesii.org>, jose luis ricon <artirj at thesii.org>

Alright, me and Jacky had the wonderful idea last night of taking the work
being done to integrate Sphinx into Wintermute, and thought about releasing
it as a seperate application, a la Dragon NaturallySpeaking.

Sphinx itself is almost complete (though it's accuracy needs some work) and
turning Sphinx into an end-user application where such users can install
from a .deb file and start dictating their documents requires around three
developers working for around five months to produce an interface for it.

1. The interface must take speech input and pass it to the system as
keyboard input.

2. The interface must be able to provide a clear, simple and above all,
easy training system (we could throw in a few Creative Commons short
stories from feedbooks for the system to learn how to  recognize it's users
voice )

3. The system must have a large (and I mean VERY large) dictionary to match
speech to words.

4. The system should also have a grammar and punctuation checker/correcter
(so users wont have to say 'Period' at the end of every sentence) This
would also be handy for Wintermute, and thankfully there are a large number
of high quality systems under GPL licensing.

5. The system must be able to accept user-defined words and record the
users saying it.

6. The system must be not only open-source, but fully extensible; other
applications/games could choose to integrate with the system

7. It must be able to check microphone quailty (by getting the user to read
out a pre-wrrtten sentence) and also check volume (so it can hear the user)
and input.

8. The system could work with the VoxForge project, and by opting in, our
users could send soundbytes (along with the corrected output) to extend
F/OSS work into speech synth and recognition with no cost to themselves.

9. Must accept certain keywords to either listen to microphone input and
ignore it.

That about covers it;  this should plug a major hole in F/OSS, it's fairly
easy to do and it's a very realistic goal.
Two things, though;

Jacky suggested we make use of Google's Speech Recognition instead of
Sphinx; well, 1. They don't allow it, and 2. It's proprietary; I'd rather
like to keep this as F/OSS as much as possible. The other thing is that of
a name; Jacky suggested the Speech Encapsulator; but I doubt many will
instantly see the link between encapsulation and speech
recognition...thoughts, anyone?

-Danté Ashton
CPU of the S.I.I
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/ubuntu-accessibility/attachments/20111204/b4e31437/attachment.html>

More information about the Ubuntu-accessibility mailing list