RE: PC Speech Recognition

["Phil Harris" <phil@xxxxxxxxxxx>]

> as part of our set-up, some VR would be useful ... as when
> hands are full or mucky, when switches are on the other side
> of the room, and when simple acknowledgements are required ...
>
> Looking into it, though, identifies several issues likely to
> limit its effectiveness

Looking into it how deeply?

> - the vast range of people's accents
> & manners of speaking;

Which - in the case of strong regional accents - can be got around by
teaching the applications initially assuming that they support such
functionality.

> the complications of room acoustics & noise levels;
> hardware fidelity / quality;

Which can be got around on the whole by use of appropriate microphones and
audio processing hardware before it hits the PC...

> and our very limited understanding of how
> language actually works & words are actually used

Are you saying that we don't understand how language works? I think we
generally do as we use it every day quite sucessfully.

I assume you mean the ability to translate our own abilities to manipulate
and use language into computer programs? In that case then there isn't a
need for computers to be able to process free form language fully in this
instance. For VR in a HA environment you want to be sure that you
"wake up"
or otherwise prepare your system for a command with a known phrase or sound
/ action and then give a brief and unambiguous instruction.

I wouldn't want a VR system to be randomly "listening in" to what
was going
on in the house (or perhaps what was being said in a TV program) and trying
to understand it and what parts of it might be sensible as instructions for
it ... You generally want the range of understanding of any VR system to be
tightly controlled and that makes the software writers job much easier too!

> ...  all in
> all, there's still a long way to go !

Not as far as you may believe ... However it might be a while before you
can
afford to do the job properly for the kind of budget you're looking at if
your target market is the IBM ViaVoice ballpark.

> We're going for Idratek,

Really? I'd never have known... ;-)

> and VR is an option on offer ...
> perhaps with IBM ViaVoice, although I've never heard of a
> user who doen't express a lot of reservations about its
> capabilities, presumably because of the very same issues ...

Perhaps find out the issues they have rather than presuming what they might
be?

>  From what I've seen of regular computer sound,

OK - I won't take the p*ss here...

> the quality
> of the hardware leaves a lot to be desired ...

What do you mean by "the quality of the hardware leaves a lot to be
desired"?

The are some excellent audio output cards available for PC if you need high
quality output and in the case of input (which is more relevent here) then
there are all manner of quality solutions at prices ranging from the merely
eyebrow-lifting to the criminally insane. If you wish you can even do all
the analogue to digital and digital to analogue conversion away from the PC
and simply feed bitsreams of audio in or out so I don't see what is to be
desired - perhaps you simply mean at the budget you are willing to spend?

> and yet we
> know from our own experience that when sounds are clear they
> are much easier to hear & understand, even if they are quiet
> ... the same must surely go for VR

Yes ... If you can feed a PC running a VR solution a good clean audio
source
then it will work better.

> !   BTW, I've never heard of VR using stereo, and yet we know
> it makes a
> big difference to our own ability to hear & understand sounds ...

No - hearing in stereo does *NOT* make "a big difference to our own
ability
to hear and understand sounds" - it allows us to *LOCATE* the source
of
sounds more easily and hence, in the past, know which direction lunch (or
something that we might be lunch for) is coming from or nowadays know where
the truck is even though we can't see it or enjoy the sound of being in the
middle of a movie gunbattle.

Feeding a VR system a stereo source will open up a whole new can of worms
as
then your VR stystem will be presented with much more in the way of echoes
and out of sync audio to deal with and process out...

Phil