Re: Observation about group

[mjmahon@aol.com (Michael J. Mahon)]

>Jeff Blakeney wrote:
>
>> On Fri, 14 Dec 2001 17:21:47 -0600, Roy Miller <millers@inetnebr.com>
>> wrote:
>>
>> >I've often wondered why a computer couldn't have a speech recognition card
>that
>> >was in essence a second computer dedicated solely to interpreting the
>speech. IF
>> >the total resources of, say, a pentium 100 were devoted to speech
>recognition,
>> >could it do the job? And then send the code to the OS of the host computer
>to
>> >perform the task commanded?
>>
>> That's exactly how the speech recognition card that Steve Circia
>> designed and published in Byte magazine in his column Circia's Circuit
>> Cellar worked.  I've always wanted to track down that article and
>> build one of those cards and maybe expand on it.  It is a very simple
>> design in that you simply record yourself saying something and it
>> compares whatever it hears with what it has recorded.  It could record
>> up to 64 samples and when it found a match it would send a single
>> character or string to the Apple II so you could say catalog and it
>> would send the text "CATALOG" followed by carriage return to the Apple
>> II.

There is a small (about 2" x 2") circuit board now available from several
"hobby" sources that contains a DSP that does speech recognition.

It is easily trainable, and delivers up to 32 "recognition signals" upon
hearing a recognized word or phrase.  It can even work in "wakeup" mode,
in which it listens to everything, waiting for a "wakeup" word or phrase to
arm it for further recognition.  It costs about $50.

You might want to check into it.


-michael

 Email:  mjmahon@aol.com
 Home page:  http://members.aol.com/MJMahon/