Hi there, I plan to write my very own voice recognition engine but i do not know how to start writing one. I know it requires biometric. Wow.. that sounds pretty advanced topic. Does voice recognition uses neural networks like training purposes? I think voice recognition need to identify sound waves... Any idea? Is there any sample app which i can refer to? At least the source code provided! Any help? Thanks. Regards, Chua Wen Ching

quote: Original post by wenching
I plan to write my very own voice recognition engine but i do not know how to start writing one. I know it requires biometric. Wow.. that sounds pretty advanced topic. Does voice recognition uses neural networks like training purposes?

"Voice recognition" covers many very different tasks. Do you want you system to identify individual speakers (as in a security application) or understand what they say? How many distinct speakers does the system need to handle? If more than one, how consistent are these speakers (gender, age, accent, etc.)? How large a vocabulary is needed? Does the vocabulary need to be extensible? What level of accuracy is required? Does the system need to operate in real time?

You will need to answer these questions before you can design an appropriate solution. Some voice recognition systems are very complex, with all sorts of spectral pre-processing and neural networks, etc., while others are relatively simple. NASA built a system allowing the voice input of the ten numerals (0 - 9), using only a simple binning system and fuzzy logic.

Sounds like it is harder than i think... Haha!

Hmm.. well i want to implement voice recognition in my FPS coming game..

There is a console which i am the security guard, then i will use my voice to open up that door...

I only need to say "Fire" (well these is speech recog)

but i want it to validate only my voice..

I not sure about real time.. is this real time?

Hmm.. i really need more info to get me started.. fuzzy logic itself is a large topic.. no kidding..

just how to get started to play with sounds validations.. hmm? i had no idea.. any help, please?

Regards,
Chua Wen Ching

You want the computer to recognize the voice of a single player, to authenticate his identity? Are you trying to prevent other players from pretending that they are that player? Or do you just want the player to be able to issue voice commands ("turn left", "fire", "jump")? The more specifics you can provide, the better.

quote: Original post by wenching
...well i want to implement voice recognition in my FPS coming game..

There is a console which i am the security guard, then i will use my voice to open up that door...

I only need to say "Fire" (well these is speech recog)

but i want it to validate only my voice..

I not sure about real time.. is this real time?

Hmm.. i really need more info to get me started.. fuzzy logic itself is a large topic.. no kidding..

just how to get started to play with sounds validations.. hmm? i had no idea.. any help, please?

I am more into "Are you trying to prevent other players from pretending that they are that player?"

Any help?

Thanks.

Regards,
Chua Wen Ching

I would like to know more about this too. If anyone has some time to share it's knowledge I would appreciate if he writes some words

EDIT: One of the first matches I've got when googleing

[edited by - xaxa on August 30, 2003 4:47:38 AM]

I''m working asr( automatic speech recognition). I have learn it at school. If you want to get seriously into it, I recommend you to read this book , for nom, I think it''s only in french : "Traitement de la parole, Presses Polytechniques et Universitaires Romandes" (It''s from my teacher, he invented a very known algorithm for speech synthesis).

I want to tell you that speech recognition is a very difficult process. You have to read a lot of articles before you start. Try search in google. you can watch on www.multitel.be, we sell speech recogntion solution, on www.babeltech.be too


The last solution is to bought a Mac with Mac OSX who has a built-in speech recogntion engine :-)

quote: Original post by wenching
I am more into "Are you trying to prevent other players from pretending that they are that player?"

Any help?

A place to start would be to use a password or passphrase that is digitized and somehow summarized (Fourier transform, etc.). The summarized version of the data would be compared to the stored pattern for the authorized user for authentication. This could be very simple or very complex, depending on what else is required.


I am trying to understand the purpose of this in a game, though: you are trying to prevent what? ...other players from getting on a computer or network who are pretending to be the real player? Would they sneak over to his house and pretend to be him to mess up his game?

I think he wants to activate some kind of built in cheat or debugger by the sound of HIS voice, so that other people playing could not use that feature.

Thanks crazyjul, but isn''t that non english?

I don''t understand non english stuff..

yeah, quite true something like that... enabling something...

Well, any idea how to do it?

Regards,
Chua Wen Ching