Re: [Algorithms] Speech recognition methods

[Matt J <mjo...@gm...>]

Time it takes to talk about writing your own speech recogonition
software - 10 minutes

Time it takes to write a program to display flash cards with the hope
of detecting speech - 4 hours

Time it takes a team of Ph. D's from one of many award winning
univerisities to develop an effective speech recognition package with
fantastic viseme detection and well implemented yet fast Hidden Markov
models - limitless

There are some things persistence can't buy, for everything else...
there is.... another mailing list :-)

Matthew

> Has anyone here ever tried to do speech recognition? I'm currently trying to
> make a little game for my kids, and it's proving to be a lot more work than
> I intended! I wanted the game to include reading flashcards, and to use some
> SR system to test them. But at the moment I'm stuck with tying into
> Microsoft's Speech API, which (a) isn't very good and (b) is a lot more
> OS-level than I usually like to go (Dammit, Jim, I'm a Director developer,
> not a magician). So I'm currently thinking about trying a FFT + genetic
> algorithm / neural net approach. (I have no idea what system the SAPI uses,
> quite probably it's something along these lines itself). I'm looking for a
> 'dictation' (sound input -> phoneme / word guess) rather than 'command and
> control' (sound input -> nearest fit word from rule) approach.
>
> Anyone got any tips or links I could try? Preferably something conceptual
> that won't leave me wallowing in a morass of long equations...
>
>