Main > Main Forum

Ultimate Arcade III / Speech recognition

<< < (2/2)

)p(:

--- Quote ---I Long story. And third, I want to have audible feedback to commands, which GC supports, but I want the feedback to be generated using text-to-speech, so it can include the name of the game it's gonna play.
--- End quote ---


I use the ms agent technology in Emulaxian to do the text to speech...and for fun did some experiments with speech recognition...I must say it was very fun especially shooting the galaxians...left left left...shoot  :D

Peter

Dudah:
Would that happen to be Micro$oft Speech SDK? My friend Aaroncake who made an MP3 computer for his car has that on it. his website is http://www.aaroncake.net I believe if you wanna check it out. Best of luck!

-Dudah

RacerX:
I have played with this too.  I wrote a little front-end in Visual Basic using Microsoft's speech recognition.  I was able to say, "Play Donkey-Kong" and it would fire off the game.  I wanted to find a front-end that was already written and just add this code to it (I'm lazy that way  ;)), but I gave up trying to find one written in VB that was open source.

1UP:
What sucks about voice input is all the damn setting up before you can even use it.

Jonathan_the_Red:
1UP--

Yeah, I wouldn't want to use speech in a game, either (apart from being slow and inaccurate, it also strikes me as cheating). I'm much more interested in it for the front end.

Microsoft operating systems have a built-in speech subsystem, and separate vendors can plug in their own recognition engines. For each user, it maintains a "speech profile". The practical upshot of this is that you can spend a few minutes training, and your speech recognition will be improved over all apps that use it.

Not only that, but training often isn't even necessary. There are two main types of speech recognition: dictation and voice commands. With dictation, you're speaking text and the computer is translating it into the written word. Dictation is a very, very computationally difficult problem, and a lot of training is required to get anywhere near acceptable accuracy. Voice commands are a much simpler problem. With voice commands, there's a relatively small list of valid words, and the recognition engine is just trying to decide if you've said one of them. Modern speech recognition engines have advanced to the point where they can do quite well at voice commands without any training at all.

If you have Windows XP, try installing the Plus! XP pack and firing up Windows Media Player. If you plug in a microphone, you can say "play artist Pink Floyd" and the recognition will be good enough without any training to do just that.

By the way, it's really unfair to compare your computer to your remote control. Your computer is a behemoth that runs a cycle in a billionth of a second and has hundreds of millions of bytes of RAM. Your remote control is a stupid little logic controller with virtually no memory. You'd darned well better believe your computer will do a better job at speech recognition.

Navigation

[0] Message Index

[*] Previous page