There were several requests on the forum for a Speech Recognition DLL. Today I finished it and it is working properly. Download can be found on my site:

http://www.jofre.de/?p=197

The dialog box at the beginning will be removed soon.

Oh thats i did request it, let see what can be done
Thanks man!

hey @Padmalcom,
the variable is_listening(var) when i start the program, its 1. After speeaking something it changes to 1 then 0. but its not noticed because its already one, so you just see it go zero. (this is the problem)

And each time then i speak it goes 1 ( if something is heard) and then changes to zero... i guess there is a problem there, on the initial variable value os is_listening(var)...

As its seems, when i speek is_listening(var) is 1 , when the word is said, after that variable returns to 0.

So i think you should make the initial value of is_listening(var) to zero on the dll? because nothing was said yet?


But sometimes the variable just goes 1 , if you something weird, that is not understood. hum..

Well i solved the problem..

The thing is that you call cpy mysentence everyframe, so is i compare the sentente and do a command, it will loop the command.. So i had to make a way to compare the sentence with a previouse one, and its equal, then do nothing, if its new, overwrite the last sentence and do something corresponding to that sentense.


Did you thought about implementing the add word, for a stricted grammar and not the full dictionary to improve the recognition results?

cheers

Hey I don't really understand how a limited grammar should improve the speech recognition... I had a look at the "create own grammar"-interface and creating an own one will result in a lot of work!

What you can do is to practise with your speech recognition api... that is a standard windows program that can be found in the system control.

hello padmalcom. but when you did your win 32 appl, you had tha function add word. and the program would select from the list only, the most likely to be the word you said. or you were working with the full dictionary grammar?

It improves alot! i tell you why: istead of having

that
the
this

if the engine contains only the word this.. when you say that or the, it will output only this, because its the most likely word comparing the the grammar it contains.. So it will only recognize words you added.

So with t he current grammar, if i say give me that, it can output alot of variations, like:
did that, did me that, eve me at .. and this wont result in any useful command, so by restricting the grammar to the words give, me, that, and other, i can say that command and it will be 100% a match

On the api, there is a section, where you can load no grammar, and then add word , i think its add sequence pr something..

hum.. but if you think its too difficult.

But anyway , the win32 demo you had on your webpage, is full dictionary or the grammar are just the commands on the list? i seem to be restricted grammar to me..
cheers

Apart from that great work.

Ah thats cool, didn't know that! Well maybe next week when I find a free time slot I'll have another look at the functionality.

In my Win32-app I use the full grammar! Hence the recognition is not good at all if the programm isn't trained

Is there a english version?

Firelord, this is a English version.. i wonder though if its possible to implement other languages. :X but i guess that has to do with the SDk that is englisgh.

@PadmalcomWell anyway your win32 worked very very well then! thats why i though it was a restricted grammar..

usually grammar is loaded via XML, but, i think, 5.1 has already the option to add a specifc word.. restricting the recognition dictionary, and thus, low variation..

But well thats like you said, when you got time. if you plan give de source code, maybe i could learn from it? since its something i ever wondered, the only think i could do is the engine SPEAK text to voice. using that SDK, the voice part i tried but i fail. :P