I have built the grammar i want to use, but when speaking it catches things
i say and understands it like the commands i want to recognize. So i thought,
if i have some words on my grammar, sound-alike the ones i already have, the
SRE will recognize them and i'll treat it as garbage. Does this makes sense?
Because since i started reading about speech recognition , one of the rules
to me was, smaller grammars, better accuracy, and this seems to go totally
against this.
Anyone to share some tips, experiences is more than welcome!
Smaller grammar, yes, better accuracy. The trick is to develop trigger
command words that do not sound like any others.
Marty Markoe, eMicrophones, Inc.
http://www.mymsspeech.com
will try that solution.
"mmarkoe" wrote:
> .
>