STT--Speech To Text

[H4LF82]

Ok correct me if I am wrong, but...

There is currently NO SUCH ANIMAL as a STT .pet for puppy. Am I Right?

I have been looking for about a year now....I watch the forum...occasionally I get excited when someone mentions it...

http://www.murga-linux.com/puppy/viewto ... 53&t=78556

http://www.murga-linux.com/puppy/viewto ... 45&t=81779

...but it always ends up going nowhere and the thread always dies.

Well, not this time.

You see, I am slowly going blind. Lupus, Stills Disease, and JRA have left me photo-phobic and unable to leave the house during daylight hours; I must wear DARK sunglasses indoors to see anything....and I envision a near future where I cannot see even in the darkest of rooms no matter how dark my glasses are. The ability to speak to my computer would be a GODSEND

and since I use Puppy almost exclusively now, I am committed to the cause.

I am ready to get my hands dirty, and since this has been on my mind for over a year now, I have some ideas on how one "could" go about building STT on puppy, but I will need this forums help, as I am a rank amateur with regard to programming in bash. I am not a techno-phobe however, and I speak other programming languages well enough, just not bash.

Would someone like to help, or point me in the right direction?

My idea is...to record the spoken input as a wav file, then output to a wave-signature-image (in PNG format), to use the image of the wave-signature to create a base64 string which can be stored as a literal string for comparison, then to build a database of each spoken word (with multiple cases of each recorded word wave-signature base64 string to compare against). Then the user simply spends three weeks or so recording every possible word or phrase they might speak to the STT to build the database....

I know its not elegant, and probably will not work like I hope it will, but its the best I have got.

Any thoughts? or am I just too blind to find the STT .pet file and if so will someone kindly point me to it?

because either I am downloading the pet file, or else I am beginning the process of building it myself. Today.

IF THIS THREAD DIES before I find my STT solution, I will likely never have the opportunity again to invest this much time and effort because I wont be able to see any longer.

Does anybody want to help one guy (and by proxy, a whole population of visually impaired folks) out?


Any thoughts, ideas, or opinions would be welcomed and appreciated!
.

[smokey01]

Have you tried Simon?
http://simon-listens.blogspot.co.at/201 ... n-040.html

I haven't looked at it yet but it sounds ok. I'm happy to try compiling it if it's suitable.

http://download.kde.org/stable/simon/0. ... .0.tar.bz2

Regards

[H4LF82]

I have not tried Simon.

I have tried sphinx which seems to be a part of Simon, but without much luck.

I would, of course, appreciate anything you could do. If you believe compiling it is an option and would not take too much effort on your part, then yes please and thank you. There is a BIG need for something like this in the visually impaired community, and you would have more than my gratitude if you did...compile it into a puppy pkg that is.

Cheers!

[H4LF82]

I have downloaded the simon source and un-tar-ed it.

The readme file says...

Please compile Simon with the provided build.sh script.

For full functionality Julius (http://julius.sourceforge.jp/en_index.php), the
HTK (http://htk.eng.cam.ac.uk) or SPHINX (http://cmusphinx.sourceforge.net/) may be required.

Please refer to the manual for more information.

...so I would assume that

http://julius.sourceforge.jp/en_index.php

http://htk.eng.cam.ac.uk

and/or

http://cmusphinx.sourceforge.net/

may also be necessary?

Looks promising though, certainly!

[smokey01]

It may have a bucket load of dependencies but that can be overcome.

It also looks like it might be a kde package which might complicate things. What is your preferred Puppy? Selecting one that has kde available might reduce the pain.

Cheers

[H4LF82]

right now I am using lucid 5.2.8

[H4LF82]

currently downloading trinity KDE from here....

http://www.puppylinuxjosejp2424.com/201 ... puppy.html

Let me know if I am getting ahead of myself..and thank you!![/b]

[amigo]

simon i an enviroenment which lets you run your desktopü using speech commands. The actual speech-to-text conversion is done by either sphinx or julius.

I have found julius to do a better job at interpreting speech than sphinx.

OP- you might want to look into using the version of knoppix which has speech support. Klaus Knoppers' wife is blind and together they have developed a full version with very good STT support. Trying to achieve that level of support for Puppy is next-to-impossible.

[ally]

may wish to look for flite also

it was used in puppeee and while back......

Geoffrey has written about it here:

http://murga-linux.com/puppy/viewtopic.php?t=66907

[H4LF82]

Ahh yes. flite.

Thank you ally, but flite is TTS if I am not mistaken, not STT.

TTS=Text-To-Speech
STT Speech-To-Text.

There is a difference. Fortunately there are MANY TTS options in puppy (which is GREAT when I want to hear what something says and cannot see to read it myself), and this is where the confusion comes in. Common sense says that TTS would/should also have STT, but sadly this is not the case.

I suppose that people assume they are the same (because its what common sense says) and then they dont give it much more thought, which is where the problem begins. The differences are subtle, with one doing the opposite of the other, but the subtleties have an ocean between them.

TTS reads aloud to the user, and STT hears what the user says and types it for them.

unfortunately there are few if any STT options. If I want my computer to hear what I am saying and type it for me (as opposed to me typing it)---in puppy, I am out of luck.

Windows has STT--for a FEE!.

Even APPLE has STT. If you want the headache that comes along with that...

But in Puppy... If I cannot see the blinking cursor on my screen, if I cannot see the icon to click on to open my web browser, if I cannot enter the address I wish to navigate into the address bar because I cannot find it, what other option do I have?

Herein lies my problem.

Thank you for your kind reply though ally!! I am grateful, and like you, I wish flite and other TTS solutions held more promise for me and others with visual impairments like me...and they DO, for things like reading text. Writing text tho, STT, not so much.

Cheers!

[smokey01]

I had a go at compiling julius and that worked fine. Now trying to figure out how to use it is the next problem. There are a lot of options.

[H4LF82]

THANK YOU amigo!!

knoppix sounds promising, I must admit. I have heard of it before, and I do not discount it. And perhaps you are right...knoppix may be a better solution for so many reasons...especially given that the developer designed it for his visually impaired wife. Its hard to beat "made with love especially for you..."

I do not question, or doubt, or disagree with you even a tiny bit on that point..

But ....

Trying to achieve that level of support for Puppy is next-to-impossible.

...impossible? Really?

Again, you may be right, but at this stage in the game, I am too old (and getting too blind) to consider changing teams now. For me, its the bottom of the 9th, and my bases are loaded. If I cannot get this ironed out in Puppy, then I am beating a dying horse, and woe is me. So be it.

But honestly amigo, I have SO MUCH confidence in the puppy OS for the disabled its not even funny. I first discovered it when I began rebuilding computers and giving them away on craigslist for free to the handicapped and elderly (In 2006-2007, I rebuilt and gave away over 1000 machines over the course of 18 months--ALL running puppy and ALL without charge). This operating system has come SO FAR in that time...

I call my puppy "Fido", which is latin for "faithful".

so if my choices are "built it for the wife" VS "in loving memory of my dog"...

Yeah...I think im stickin with the dog.

The wife has let me down in the past, unlike the dog.

[H4LF82]

smokey01, you my friend, are AWESOME!



THANK YOU!!

[ally]

ha ha - my bad, need more sleep............

[H4LF82]

ha ha - my bad, need more sleep............

Dont we all, ally!

no worries mate. Cheers!

[partsman]

I too have a member of my family that is blind this would be a wonderful
breakthrough for puppy ! I would love to help in any way that i can ! although I am still very new to linux maybe i could help with testing ! please feel free to ask ! I will be watching this thred ! Thanks for bringing
this to the puppy community my friend

[H4LF82]

indeed partsman!

You and I BOTH are of one mind here. A breakthrough it would be, Puppy has needed STT for far too long IMHO. Personally, I think a "Seeing-Eye-Puppy" OS developed specifically FOR the blind would go a LONG way towards helping the "Puppy Linux OS" brand...to say nothing of the benefit to society (blind and sighted) on the whole.

Maybe I will write an email to BK and say as much. I know I for one would be thrilled to be the guinea pig on that particular project.

Stay tuned...our savior today is smokey01 and he seems to have the mustard and the minerals to make STT happen as a .pet., which is a good first step in the right direction. Please save any gratitude you feel for him, for if anyone deserves to be thanked, it is surely smokey01.

[partsman]

I could not agree more smokey01 has done alot for the puppy community
not to mention hosting a website for us and our developers ! THANK YOU ! SMOKEY01 as you are yet another valuable asset to the puppy project !

[Announcer]

Yes, most likely you can install the adriane menu and utilities, which
are shell-based, on other distros, too.

Just the screenreader/braille/speech combination is preconfigured on
Knoppix, you would need to get this working on a different distros
first, then you can install the adriane scripts to run from init
on the text console by these lines in /etc/inittab:

2respawn:su -l -c "/usr/bin/adriane" knoppix >/dev/tty2 </dev/tty2 2>&1
3respawn:su -l -c "/usr/bin/adriane" knoppix >/dev/tty3 </dev/tty3 2>&1
4respawn:su -l -c "/usr/bin/adriane" knoppix >/dev/tty4 </dev/tty4 2>&1
...

(Instead of "knoppix", write the account name you would like to have as
default adriane user).

You can find the packages and sources here:

http://debian-knoppix.alioth.debian.org ... s/adriane/

Regards
-Klaus Knopper

[H4LF82]

Thank you Announcer, but again, no thank you.

Im not turning my back on Puppy now just to go over to Knoppix and start the whole learning curve all over again--I just cannot bring myself to. Ive had YEARS to learn how to use puppy and I have grown comfortable with it...despite its shortcomings (of which there are few and they are far between). Speech-to-text and the lack thereof is my one and only tiny complaint worthy of being mentioned...

Smokey01 is getting the simon package compiled for me even as I type this, and Ive got to tell you, I have the faith of a mustard seed in his abilities. He seems to think it is do-able and he should know.

He compiles software in his sleep.

So when smokey01 says "Forget it man. Get Knoppix or else suck it up and deal!", THEN I might consider it.

But at this point, you might as well be telling me to switch back to windows "because it has Nuance Dragon Speaking".