When are we gonna get voice recognition for computers?

[frenchn00b]

Here what we could program:
http://www.youtube.com/watch?v=gptzXomX-GQ&feature=related
;)


wikipedia: voice recognition (http://en.wikipedia.org/wiki/Speech_recognition)

[User3k]

I am worried if my computer heard me half the time, especially when I am trying to get something to work that should work but doesn't, then it would choose not to boot.

[keplerspeed]

Same as flying cars and hoverboards. Give it a good 50 to 100 years, then double your estimate.

PS: Its like cooking bacon. Always double what you think is enough.

[3rdalbum]

The Mac OS has had voice recognition for ages. In fact, Linux did have some degree of voice control a little while ago; but I don't think it came to anything.

[AllRadioisDead]

Am I the only one that would find it weird talking to a monitor?

[frenchn00b]

The Mac OS has had voice recognition for ages. In fact, Linux did have some degree of voice control a little while ago; but I don't think it came to anything.

there is a new binary opensource made for ubuntu working:
http://sourceforge.jp/projects/julius/downloads/44145/julius-4.1.3-linuxbin.tar.gz
http://julius.sourceforge.jp/en_index.php
it works not that bad

[User3k]

Am I the only one that would find it weird talking to a monitor?

I would find it stranger if the monitor was talking to me. Can you picture running Windows and it arguing with you because it is always right? The future looks dark right now, lol.

[lisati]

Vista on my laptop came with some kind of voice recognition. I've never used it.

[keplerspeed]

IT crowd anyone?

[frenchn00b]

I would find it stranger if the monitor was talking to me. Can you picture running Windows and it arguing with you because it is always right? The future looks dark right now, lol.

your are right, if it sponsored by some billy g.
But if it is Linux, that's brighter


edit:mr bean (http://www.youtube.com/watch?v=vdcCxiqCNC0&feature=fvw)

[ssam]

IT crowd anyone?

:-)

[Exodist]

Here what we could program:
http://www.youtube.com/watch?v=gptzXomX-GQ&feature=related
;)


wikipedia: voice recognition (http://en.wikipedia.org/wiki/Speech_recognition)

Most operating systems have voice recognition. I think MacOSX has the best currently from what I have seen even a few years back. But others could have stepped up better since then. So I didnt vote..

[clanky]

My phone (nokia E71) has voice recognition, I don't use it though because I feel silly talking to my phone when no-one is at the other end.

[Exodist]

IT crowd anyone?


most epic video of them all.. (http://www.youtube.com/watch?v=QAUyaELfwBo&NR=1)

[keplerspeed]

most epic video of them all.. (http://www.youtube.com/watch?v=QAUyaELfwBo&NR=1)

Awesome. Haven't seen that series yet. I was referring to a scene is the first or second seasons, "hello computer!".

[Exodist]

Awesome. Haven't seen that series yet. I was referring to a scene is the first or second seasons, "hello computer!".
LOL yea, that one reminds me of Scotty off of Startrek IV: The voyage home (http://www.youtube.com/watch?v=wzRziK-kZtQ)

[keplerspeed]

So that's where they got the joke from... all this intertextuality that I dont pick up on...

[User3k]

While we already have voice recognition. I am thinking more along the lines of Knight Rider. Where I could actually have a conversation with my computer (car, phone, spaceship.)

I am not sure I want my computer talking (Or thinking) But on the other hand, that could be interesting... I can see it yelling at me when it needed an upgrade. :D

[frenchn00b]

http://www.youtube.com/watch?v=QAUyaELfwBo&NR=1 Awesome. Haven't seen that series yet. I was referring to a scene is the first or second seasons, "hello computer!".

Indeed. very funny

[Psumi]

Even though we do not need it to be honest. Especially for laptops. Kind of weird talking to your laptop while other people look at you.

[frenchn00b]

Even though we do not need it to be honest. Especially for laptops. Kind of weird talking to your laptop while other people look at you.

Additionally, if it is like vista, one will need 20 CPU + ram of 20 billions mega bytes + a high tech microphone

well, and then cpu 100pct all the time, great for burning ;)

that's called future of voice recognition. So please, sure not before 10-20 years, or 50 years

[Dr. C]

How about 1996? Here is a web page from that era discussing voice recognition on various platforms including for example OS/2 Warp and even DOS.

http://www.voicerecognition.com/voice_article_vol_2.html

[AICollector]

Voice Recognition is a lovely technology; as a writer I find it SO much easier to talk then to type...

That's about the only thing I miss about Windows...

Anyway;

Voice Recog isnt...well, easy. You need an acoustic model (to filter out voice from sounds) and a langauge model (to fill out words from voice)

We've got CMUSphinx which contains an acoustic model (it can detect when a human is speaking) but no language model (it cant understand what you say)

Also bear in mind language, accent, inflection, illness, fun with helium balloons....all require their own language models (and in some cases, a diffrent acoustic model) you can't just have an English model and make it understand French, it has to know the syntax (We say: A Blue Coat, the French say: A Coat Blue)

That's a LOT of local work.

and let's not forget we need loads of computing power to make it understand (IE: Best close everything else down) and about five minutes worth of training generates some 500mb of data for the program....

I would love it if it came soon, very soon. I know plenty of disabled folk who cannot use Linux because they are dependant upon voice input :(

[frenchn00b]

Voice Recognition is a lovely technology; as a writer I find it SO much easier to talk then to type...

That's about the only thing I miss about Windows...

Anyway;

Voice Recog isnt...well, easy. You need an acoustic model (to filter out voice from sounds) and a langauge model (to fill out words from voice)

We've got CMUSphinx which contains an acoustic model (it can detect when a human is speaking) but no language model (it cant understand what you say)

Also bear in mind language, accent, inflection, illness, fun with helium balloons....all require their own language models (and in some cases, a diffrent acoustic model) you can't just have an English model and make it understand French, it has to know the syntax (We say: A Blue Coat, the French say: A Coat Blue)

That's a LOT of local work.

and let's not forget we need loads of computing power to make it understand (IE: Best close everything else down) and about five minutes worth of training generates some 500mb of data for the program....

I would love it if it came soon, very soon. I know plenty of disabled folk who cannot use Linux because they are dependant upon voice input :(

I does really work, DNS? which version you have, and with hardware? In english? It is usually said that it makes a lot lot of errors DNS

and the problem with DNS, is that you cannot configure it, to run commands
nothing more handy, that asking you pc some pre-programmed sentences, like perl-sphinx voice command.

[sandyd]

http://www.linux.com/archive/feature/41723
^ ^
this

[AICollector]

http://www.linux.com/archive/feature/41723
^ ^
this

Ahhh, wrong type, my dear.

That's a closed domain system, it cant learn new words nor can it understand natural langauge input (though those systems have an accuracy rating of 99% or so)

What we're talking about here is an open-domain system, one that can do the exact opposite as well as a closed domain systems tasks.

[markp1989]

i had a voice recognition program on one of my old pcs, dragon dictate i think it was called.

it wasnt very good , took along time to calibrate it all to my voice, and in the end just got frustrating

alot of phones have voice reconignition for dialing, been on them since like 2000 , but i wouldnt want to use it as my main input .

[AICollector]

Dragon Dictate evolved into Dragon NaturallySpeaking.

Now at version 11, it's perhaps the most advanced speech recog. system available.


and no, it dosent work in Linux :( (though it does have a silver rating with WINE, most of the important functions are broken)


If I could make one wish, just one, it would be to make Dragon open-source...


but no :/

[oldos2er]

I had speech recognition software in 1996, with OS/2 Warp 4.0, and it worked extremely well. It had two distinct functions; one to navigate menus, open and close programs, etc., and a dictation mode for creating documents. The dictation function required a training period, but the menu navigation software worked out-of-the-box.

I really miss OS/2.

[AICollector]

sounds like the dictation model was open-domain, and the controller was closed-domain.

Also, interesting KDE-centric project called Simon going on that provides a voice controller.

[frenchn00b]

http://www.linux.com/archive/feature/41723
^ ^
this

those programs are hopeless. it is like talking to a wall

[Marvin666]

Vista shipped with some voice recognition stuff. I tried setting it up, and no matter how many of those things that trains it, it still didn't work right.

[frenchn00b]

Dragon Dictate evolved into Dragon NaturallySpeaking.

Now at version 11, it's perhaps the most advanced speech recog. system available.


and no, it dosent work in Linux :( (though it does have a silver rating with WINE, most of the important functions are broken)


If I could make one wish, just one, it would be to make Dragon open-source...


but no :/

it is the best one, and still have poor performances.
One need a super machine, tons of rams, ultra calibrating, ... ultra expensive mic... and so on
it gets worse with the time even, it does: anti-learning

the best would be a machine to learn progressively your voice.
Each word after each word, not like DNS. THough, you can with DNS, but it is very poor result in any cases.

[frenchn00b]

sounds like the dictation model was open-domain, and the controller was closed-domain.

Also, interesting KDE-centric project called Simon going on that provides a voice controller.

that's so pitty that they forget a backend first. It wont bring much if it cannot be backend. Poor decision.

Linux needs command lines for servers and frontends. We arent developping like Skype.

simon for ubuntu


here the files for i386:
type this:


wget "http://mesh.dl.sourceforge.net/project/speech2text/simon/0.2/libportaudio2_19%2Bsvn20090605-0ubuntu7_i386.deb"
wget "http://kent.dl.sourceforge.net/project/speech2text/simon/0.2/simon-0.2-Linux_i386.deb"
apt-get -f install
dpkg -i "libportaudio2_19%2Bsvn20090605-0ubuntu7_i386.deb"
apt-get -f install
dpkg -i "simon-0.2-Linux_i386.deb"
apt-get -f install
dpkg -i "simon-0.2-Linux_i386.deb"



as user type:

simon

You can try this soft and let us know whehter it is good or not