David,
Accurate speech recognition is one of the most difficult tasks you can
imagine, even if it's only for "simple" commands. I'd strongly discourage
you from trying to "roll your own" system.
If it's possible to use it, even if it seems like overkill, go with the
Microsoft Speech Server framework -- it's designed for .Net, has a nice
interface and documentation, and is designed by some of the best and
brightest speech researchers. Your other options might be to to hook into
the Office XP speech recognition engine (if it's possible), or buy a
third-party speech component (.Net or ActiveX.) Check out these other
links:
http://research.microsoft.com/srg/
http://support.microsoft.com/?kbid=306901
There are also a variety of third-party speech components available -- just
go to ComponentSource.com and search for "speech."
If you're absolutely certain that you want to build your own system, there
are two recent textbooks on the topic: Jurafsky & Martin, "Speech & Language
Processing," and Allen, "Natural Language Understanding." (Both available
from Amazon.com.) However, there is no simple off-the-shelf algorithm that
you can just drop into your code and expect to work. Only go this route if
you have many months of time on your hands. <g>
Hope this helps,
Robert Jacobson
