[MV] Re: Questions on customising VV

T. Patrick Henebry tphenebry at comcast.net
Thu May 29 14:45:29 PDT 2003


>Good speech recognition and accuracy REQUIRES one to use "audio setup" on a
>regular basis (ideally each time one starts up VV).  I don't always do that.
>Accuracy always improves when I do so.

Very good point. I should not have left that out. My previous post 
may not have been exactly clear. The speech recognition engine uses 
multiple sources of data to turn speech into text. For best possible 
results the microphone settings are kept separate from the voice 
model so that the microphone settings can be adjusted on the fly to 
accommodate different conditions without affecting the voice model.

For best results you *should* run microphone setup when you first 
start ViaVoice.  Microphone setup should also be run when the 
background conditions change such as a fan being turned on or off.

If the changing background conditions significantly alter the way you 
speak then it's worth creating another voice model for those 
conditions in order to get the best results. But you will still have 
to run microphone setup. after switching voice models. One of the 
nice things about crating another voice model in ViaVoice is that any 
text macros or special words you added to the vocabulary in the first 
voice model are retained in subsequent voice models. So you don't 
have to renter them.

Different voice models for different microphones is also a good idea.
-- 
----------
T. Patrick Henebry



More information about the MacVoice mailing list