Speech Recognition and delphi

Hi Experts,

Is there any way to make a program of speech recognition with delphi. A program where I speak a word in the microphone and the program writes the word in a memo.

this is possible?

Thx for some help
Alexandre
alpiresAsked:
Who is Participating?
 
MerijnBConnect With a Mentor Sr. Software EngineerCommented:
Do you have any background in this (specialized) kind of work? Is there a reason you want to do it yourself and can't use one of the well known products (i.e. naturally speaking)
0
 
JosephGloszConnect With a Mentor Commented:
To actually answer your questions, (1) yes there is certainly a way to do speech recognition in delphi. Delphi   can certainly handle it as well as C++ or C#.  To your second question, yes it is very possible.

But, and I think this is what MerjinB might be getting at, to even ask this question implies you know little about the subject matter. Maybe you are looking for pre-built component or class of some kind?

If this is the case, then this job will be a very large undertaking. Maybe that's OK though because you want to learn  what's involved.

And what's involved is very very big.  Language processing is one of the really amazing things humans do, and we do it lightning-quick.

For software, it is much harder. Historically there have been two main categories in this area: the speaker-dependent and speaker-independent cases. Up until just recently, speaker-dependent speech recognition was all we could do, for all practical purposes. We just didn't have the processing power.

Speaker-dependent processing is where you have to "train" the software to recognize how you say certain words and phrases.  We humans pronounce things so many different ways, how could software ever handle that?

Say "Park My Car."  Three words. Should be easy for speech recognition, right? Have a Bostonian say it, and the same waveform, which is all the software has to look at, comes out as "Pock My Cah"  Non-sensical.

But have that Bostonian train the software, and it would recognize the way he says "park" and most of the time it would work.

But even then, because of homonyms, and similar sounding phrases, we also need context. The waveforms for "Ice Cream" and "I scream" are identical.  So now the software has to see what is in the sentence, parse it somewhat, and then take a guess at what  you are saying.

And, on top of this, the software has to be able to tell where words begin and end. Up until just recently, we had to actually pause briefly between words when speaking them.   Try to just write software to turn "whachadoin'?" into "what are you doing?" to see how hard that is.

I long time ago, I tried (a horribly misguided attempt) at doing this. Later, I was tasked with integrating an existing high-end recognition package into our phone system so customers could "say" what they wanted. That didn't work either.

And, I still don't think it works all that well. Not nearly close to 100% to make it useful. At least, in my book.

But if you want to explore this, just google "delphi speech recognition"

You can start with something like this:

http://www.delphi3000.com/articles/article_2629.asp?SK=

then as you learn the million issues involved, maybe you'll develop something better...  

Good luck!

Joseph




0
 
alpiresAuthor Commented:
Thanks expets, i will use sapi 5.1
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.