 Speech ComputingWhere Speech Meets Computing |
|
|
Registering and logging in removes this ad. User loginRegistering and logging in removes this ad.
Registering and logging in removes this ad. |
HAL2000 Speech Recognition Engine
Submitted by ryokanNHC on Tue, 09/04/2007 - 18:45.
Does anyone know what speech recognition engine the home automation software (HAL2000) by automated living uses? From some reviews it seems to have fairly decent speaker independent recognition capabilities. |
Registering and logging in removes these ads.
LifeLock is the only Identity Theft Prevention Solution backed by a one-million dollar guarantee!Click here to get a 10% discount. $10 off $200 valid in our Computers, Software, Home Networking, Digital Cameras and Electronic Stores! exp 10/31/08 Registering and logging in removes these ads. MultiSite SearchNavigationS/R TipsPopular contentWho's onlineThere are currently 0 users and 6 guests online.
|
NOTICE OF DISCLAIMER: All readers, visitors and/or users of this site is/are hereby given notice that any and all either express or implied views, opinions, suggestions, advice, product promotions, and/or any apparent statements of truths or facts that have been, presently are or in the future may be placed on this site are to be understood to be solely those of the party posting same. In no regard or event are any or all express or implied views, opinions, suggestions, advice, product promotions, and/or any apparent statements of truths or facts posted on this site in any way whatsoever to be either directly or indirectly understood, interpreted, construed or relied upon by any readers, visitors and/or users of this site as being expressed or purveyed by or representative of the host of this site, nor shall same be directly or indirectly understood, interpreted, construed or relied upon by any readers as an endorsement of same by the host of this site. Any and all such express or implied views, opinions, suggestions, advice, product promotions, and any/or apparent statements of truths or facts posted on this site should be thoroughly investigated independently by any and all readers, visitors and/or users of this site prior to relying or acting upon same in any way or at any time.
© 2005-2008 Speech Computing
Best Viewed at 1024x768 and higher resolutions
Although I don't know this
Although I don't know this for sure, I do have some experience (long ago) in using voice control software.
I don't know if HAL2000 uses this, but the older stuff had something akin to pattern recognition instead of a real speech recognition engine. The old ones you had to train for each word about 3 times then it averaged the patterns.
I understand the newer ones are pretty good at independent speaker recognition but it's because they have a pretty limited vocabulary so there aren't many choices.
I'm sure others will know more. Just wanted to relate my thoughts!
Skip
Probably does have a pretty
Probably does have a pretty limited vocabulary. Its interesting how accurate systems can be if they only have a list of limited possibilities and known responses. The MS engine only has one really big dictation grammar as far as I can tell. I wonder why dictation systems ship with 60,000+ word vocabularies, seems like that would almost be a a disadvantage in the accuracy department. How often do you use the word 'malapropism' when you meant 'mail the powerpoint'
"Rue the day?" Who talks like that? ...
Real Genius reference for anyone that cares.
If you devise a feasible
If you devise a feasible strategy for determining exactly what vocabulary a particular user wants, you might a have a possible future with Nuance sales and marketing!
If you hang around SR long enough you also might figure out the advantage of shipping with a 60,000+ word vocabulary.
You might also try figuring out the marginal accuracy/recognition-time cost of adding word N+1 to a vocabulary of N words and plotting those equations. My hunch is that for moderately small N, those costs are negligible.
It is feasible for a user to develop a custom restricted vocabulary, and some advanced users do just that.
Bruce
This is an apples versus
This is an apples versus oranges issue and it doesn't really matter what speech recognition program is used in home automation because they have an extremely limited vocabulary which that is why they work so well. These types of systems also need to work well over a distance (as opposed to close micing with NaturallySpeaking) but don't think for a moment that they are capable of handling real dictation. Automation systems are similar to the information systems used by phone companies that start off asking you to say the name of the state, followed by the city, etc..
Lunis Orcutt - Developer of KnowBrainer (DNS Command Software)
Now Providing FREE (1st 5 min.) Tech Support 615-884-4558
A Nuance Gold Certified Endorsed Vendor
Not really comparing it to
Not really comparing it to anything in particular really, but ....
I think ideally speech rec would JUST be a dictation type engine. But currently in the non ideal world we are in, speech rec has alot of forms I think beyond dictation. As an extreme example, check out Jott for example. Very little computational speech rec., just send the sample to some person in India who listens and types it in!. Still an argument for it being a form of automated speech recognition.
As far as Hal2000, its more the speaker independence (over a distance) part that interests me.
I am more just curious what the engine was. I read in some postings elsewhere that it was an older version of the L&H engine. The specs given on the website indicate older machine which dates roughly to that time as well.
Wondering if anyone here had any confirmation knowledge.