->Speech recognition?3
Archiver
Posts: 46,084
I want to beable to speak into a mic with words such as "turn
on", "move left", etc... Im sure its not going to be easy but thats
what i am going to do but i need to know how first?? what would you
recomend that i get to do this? and how whould i sample the data to
find a match?
--- In basicstamps@yahoogroups.com, "George Bush" <capthoohoo@y...>
wrote:
> --- In basicstamps@yahoogroups.com, "virtualeast" <vea@v...> wrote:
> > What would i need to write/do that would be able to understand
>>what i am saying if i use pbasic and the microcontroller with
>>eeprom?
> > what would i need to do if i use a PIC?
> > and what would i need to do to understand the numbers that the
>phone is dialing with pbasic?
> > p.s. thanks, these are all good thoughts, im sure im going to ask
> > more too. =-)
> >
> The first thing you do is decide what you need as specifically as
> possible. Are you making "The Clapper" or are you trying to make
> something that would emulate the functionality of the cell phone?
I would suggest starting with The Clapper. It would give you
>practice with the elements you would need to move on to more
>complicated sould recognition. The answer to all your questions is
>in the specifications of what you need to build.
>The three blocks I talked about before could be the basis of your
>design for all of the cases you ask about above. 1) High speed A/D
>(How fast, how much data?) 2) FFT function 3) known samples in
>memory with a "best fit" comparison algorithm. In each of your
>cases these blocks get implemented differently, but conceptually you
>need all three. In Tim's phone suggestion you would be sampling
phone >tones rather than the human voice.
>
> Neil
on", "move left", etc... Im sure its not going to be easy but thats
what i am going to do but i need to know how first?? what would you
recomend that i get to do this? and how whould i sample the data to
find a match?
--- In basicstamps@yahoogroups.com, "George Bush" <capthoohoo@y...>
wrote:
> --- In basicstamps@yahoogroups.com, "virtualeast" <vea@v...> wrote:
> > What would i need to write/do that would be able to understand
>>what i am saying if i use pbasic and the microcontroller with
>>eeprom?
> > what would i need to do if i use a PIC?
> > and what would i need to do to understand the numbers that the
>phone is dialing with pbasic?
> > p.s. thanks, these are all good thoughts, im sure im going to ask
> > more too. =-)
> >
> The first thing you do is decide what you need as specifically as
> possible. Are you making "The Clapper" or are you trying to make
> something that would emulate the functionality of the cell phone?
I would suggest starting with The Clapper. It would give you
>practice with the elements you would need to move on to more
>complicated sould recognition. The answer to all your questions is
>in the specifications of what you need to build.
>The three blocks I talked about before could be the basis of your
>design for all of the cases you ask about above. 1) High speed A/D
>(How fast, how much data?) 2) FFT function 3) known samples in
>memory with a "best fit" comparison algorithm. In each of your
>cases these blocks get implemented differently, but conceptually you
>need all three. In Tim's phone suggestion you would be sampling
phone >tones rather than the human voice.
>
> Neil
Comments
alone, so why not simply use a Speech Recognition Kit
like jameco carries:
http://www.jameco.com/cgi-bin/ncommerce3/ProductDisplay?prmenbr=91&prrfnbr=35553\
3&cgrfnbr=501&ctgys=
..."A complete easy to build programmable recognition
kit, program the words you want the circuit to
recognize. The heart of the speech recognition kit is
the HM2007 IC wherein you can program 40 words @ 0.96
seconds word length or 20 words @ 1.92 seconds word
length. Several applications for this kit are: speech
controlled appliances and toys, speech assisted
computer games, telephone assistance systems, voice
recognition security, and speech-to-speech
translation. ..."
i have never used it, but i imagine upon a positive
word match it simply sets an output HI, which u can
then easily read with the stamp and then act
accordingly. seems easy enough to me, albeit a little
costly as the kit is ~$120. i did a quick search for
"speech" on digikey's site, not a lot popped up but
this looks promising:
http://www.sensoryinc.com/
they appear to have a cheaper version????? the manual
can be downloaded at the bottom of the page.
hope that helps.
ross
--- virtualeast <vea@v...> wrote:
> I want to beable to speak into a mic with words such
> as "turn
> on", "move left", etc... Im sure its not going to be
> easy but thats
> what i am going to do but i need to know how first??
> what would you
> recomend that i get to do this? and how whould i
> sample the data to
> find a match?
>
>
> --- In basicstamps@yahoogroups.com, "George Bush"
> <capthoohoo@y...>
> wrote:
> > --- In basicstamps@yahoogroups.com, "virtualeast"
> <vea@v...> wrote:
> > > What would i need to write/do that would be able
> to understand
> >>what i am saying if i use pbasic and the
> microcontroller with
> >>eeprom?
> > > what would i need to do if i use a PIC?
> > > and what would i need to do to understand the
> numbers that the
> >phone is dialing with pbasic?
> > > p.s. thanks, these are all good thoughts, im
> sure im going to ask
> > > more too. =-)
> > >
> > The first thing you do is decide what you need as
> specifically as
> > possible. Are you making "The Clapper" or are you
> trying to make
> > something that would emulate the functionality of
> the cell phone?
> I would suggest starting with The Clapper. It would
> give you
> >practice with the elements you would need to move
> on to more
> >complicated sould recognition. The answer to all
> your questions is
> >in the specifications of what you need to build.
> >The three blocks I talked about before could be the
> basis of your
> >design for all of the cases you ask about above.
> 1) High speed A/D
> >(How fast, how much data?) 2) FFT function 3)
> known samples in
> >memory with a "best fit" comparison algorithm. In
> each of your
> >cases these blocks get implemented differently, but
> conceptually you
> >need all three. In Tim's phone suggestion you would
> be sampling
> phone >tones rather than the human voice.
> >
> > Neil
>
>
> To UNSUBSCRIBE, just send mail to:
> basicstamps-unsubscribe@yahoogroups.com
> from the same email address that you subscribed.
> Text in the Subject and Body of the message will be
> ignored.
>
>
> Your use of Yahoo! Groups is subject to
> http://docs.yahoo.com/info/terms/
>
>
__________________________________
Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!
http://sbc.yahoo.com
http://www.jameco.com/Jameco/Products/ProdCT/p197.pdf
--- In basicstamps@yahoogroups.com, Klutch <klutch66@y...> wrote:
> basically there is no way to do this with the stamp
> alone, so why not simply use a Speech Recognition Kit
> like jameco carries:
> http://www.jameco.com/cgi-bin/ncommerce3/ProductDisplay?
prmenbr=91&prrfnbr=355533&cgrfnbr=501&ctgys=
>
> ..."A complete easy to build programmable recognition
> kit, program the words you want the circuit to
> recognize. The heart of the speech recognition kit is
> the HM2007 IC wherein you can program 40 words @ 0.96
> seconds word length or 20 words @ 1.92 seconds word
> length. Several applications for this kit are: speech
> controlled appliances and toys, speech assisted
> computer games, telephone assistance systems, voice
> recognition security, and speech-to-speech
> translation. ..."
>
> i have never used it, but i imagine upon a positive
> word match it simply sets an output HI, which u can
> then easily read with the stamp and then act
> accordingly. seems easy enough to me, albeit a little
> costly as the kit is ~$120. i did a quick search for
> "speech" on digikey's site, not a lot popped up but
> this looks promising:
> http://www.sensoryinc.com/
> they appear to have a cheaper version????? the manual
> can be downloaded at the bottom of the page.
> hope that helps.
> ross
>
>
> --- virtualeast <vea@v...> wrote:
> > I want to beable to speak into a mic with words such
> > as "turn
> > on", "move left", etc... Im sure its not going to be
> > easy but thats
> > what i am going to do but i need to know how first??
> > what would you
> > recomend that i get to do this? and how whould i
> > sample the data to
> > find a match?
> >
> >
> > --- In basicstamps@yahoogroups.com, "George Bush"
> > <capthoohoo@y...>
> > wrote:
> > > --- In basicstamps@yahoogroups.com, "virtualeast"
> > <vea@v...> wrote:
> > > > What would i need to write/do that would be able
> > to understand
> > >>what i am saying if i use pbasic and the
> > microcontroller with
> > >>eeprom?
> > > > what would i need to do if i use a PIC?
> > > > and what would i need to do to understand the
> > numbers that the
> > >phone is dialing with pbasic?
> > > > p.s. thanks, these are all good thoughts, im
> > sure im going to ask
> > > > more too. =-)
> > > >
> > > The first thing you do is decide what you need as
> > specifically as
> > > possible. Are you making "The Clapper" or are you
> > trying to make
> > > something that would emulate the functionality of
> > the cell phone?
> > I would suggest starting with The Clapper. It would
> > give you
> > >practice with the elements you would need to move
> > on to more
> > >complicated sould recognition. The answer to all
> > your questions is
> > >in the specifications of what you need to build.
> > >The three blocks I talked about before could be the
> > basis of your
> > >design for all of the cases you ask about above.
> > 1) High speed A/D
> > >(How fast, how much data?) 2) FFT function 3)
> > known samples in
> > >memory with a "best fit" comparison algorithm. In
> > each of your
> > >cases these blocks get implemented differently, but
> > conceptually you
> > >need all three. In Tim's phone suggestion you would
> > be sampling
> > phone >tones rather than the human voice.
> > >
> > > Neil
> >
> >
> > To UNSUBSCRIBE, just send mail to:
> > basicstamps-unsubscribe@yahoogroups.com
> > from the same email address that you subscribed.
> > Text in the Subject and Body of the message will be
> > ignored.
> >
> >
> > Your use of Yahoo! Groups is subject to
> > http://docs.yahoo.com/info/terms/
> >
> >
>
>
> __________________________________
> Do you Yahoo!?
> SBC Yahoo! DSL - Now only $29.95 per month!
> http://sbc.yahoo.com
I worked with some Speech Recognition.
Couple of tidbits...
There was a Radio Shaft chip called VCP200 (You can search on web for it).
They come up on eBay periodically. It was a chip that would have an active
pin outputs for the following recognized words: 'Left Turn', 'Turn Right',
'Forward', 'Reverse', 'Stop'. It was speaker independent so anybody could
say the words however it was not very robust and had a low percent of
accuracy.
I bought the Voice Direct™ II Speech Recognition Kit from Sensory
http://sensoryinc.com and it came with an evaluation board, speech
module, cables and software that allows you to program the module
using a programming language similar to C. I am totally satisfied with
it and bought an additional module since the only thing you need for
a project in the programmed module. It has very good accuracy and the
evaluation board microphone setup was good enough so that I could speak
from the kitchen and it would recognize what I was saying while it was in
the livingroom. A bit pricy to start off with the kit but I found it well
worth it.
I never saw the Hello-IC or ever could even find a place to order one.
I believe it would be easier to see a UFO.
After having seen the speech recognition in the cell phones and treo type
devices, I wish I could get my hands on what they are doing as it is very
impressive to me. I keep watching 8)
Bob
At 01:52 PM 7/10/2003 -0000, you wrote:
>Good suggestion -- try this link:
>http://www.jameco.com/Jameco/Products/ProdCT/p197.pdf
>
>--- In basicstamps@yahoogroups.com, Klutch <klutch66@y...> wrote:
>> basically there is no way to do this with the stamp
>> alone, so why not simply use a Speech Recognition Kit
>> like jameco carries:
>> http://www.jameco.com/cgi-bin/ncommerce3/ProductDisplay?
>prmenbr=91&prrfnbr=355533&cgrfnbr=501&ctgys=
>>
>> ..."A complete easy to build programmable recognition
>> kit, program the words you want the circuit to
>> recognize. The heart of the speech recognition kit is
>> the HM2007 IC wherein you can program 40 words @ 0.96
>> seconds word length or 20 words @ 1.92 seconds word
>> length. Several applications for this kit are: speech
>> controlled appliances and toys, speech assisted
>> computer games, telephone assistance systems, voice
>> recognition security, and speech-to-speech
>> translation. ..."
>>
>> i have never used it, but i imagine upon a positive
>> word match it simply sets an output HI, which u can
>> then easily read with the stamp and then act
>> accordingly. seems easy enough to me, albeit a little
>> costly as the kit is ~$120. i did a quick search for
>> "speech" on digikey's site, not a lot popped up but
>> this looks promising:
>> http://www.sensoryinc.com/
>> they appear to have a cheaper version????? the manual
>> can be downloaded at the bottom of the page.
>> hope that helps.
>> ross
>>
>>
>> --- virtualeast <vea@v...> wrote:
>> > I want to beable to speak into a mic with words such
>> > as "turn
>> > on", "move left", etc... Im sure its not going to be
>> > easy but thats
>> > what i am going to do but i need to know how first??
>> > what would you
>> > recomend that i get to do this? and how whould i
>> > sample the data to
>> > find a match?
>> >
>> >
>> > --- In basicstamps@yahoogroups.com, "George Bush"
>> > <capthoohoo@y...>
>> > wrote:
>> > > --- In basicstamps@yahoogroups.com, "virtualeast"
>> > <vea@v...> wrote:
>> > > > What would i need to write/do that would be able
>> > to understand
>> > >>what i am saying if i use pbasic and the
>> > microcontroller with
>> > >>eeprom?
>> > > > what would i need to do if i use a PIC?
>> > > > and what would i need to do to understand the
>> > numbers that the
>> > >phone is dialing with pbasic?
>> > > > p.s. thanks, these are all good thoughts, im
>> > sure im going to ask
>> > > > more too. =-)
>> > > >
>> > > The first thing you do is decide what you need as
>> > specifically as
>> > > possible. Are you making "The Clapper" or are you
>> > trying to make
>> > > something that would emulate the functionality of
>> > the cell phone?
>> > I would suggest starting with The Clapper. It would
>> > give you
>> > >practice with the elements you would need to move
>> > on to more
>> > >complicated sould recognition. The answer to all
>> > your questions is
>> > >in the specifications of what you need to build.
>> > >The three blocks I talked about before could be the
>> > basis of your
>> > >design for all of the cases you ask about above.
>> > 1) High speed A/D
>> > >(How fast, how much data?) 2) FFT function 3)
>> > known samples in
>> > >memory with a "best fit" comparison algorithm. In
>> > each of your
>> > >cases these blocks get implemented differently, but
>> > conceptually you
>> > >need all three. In Tim's phone suggestion you would
>> > be sampling
>> > phone >tones rather than the human voice.
>> > >
>> > > Neil
>> >
>> >
>> > To UNSUBSCRIBE, just send mail to:
>> > basicstamps-unsubscribe@yahoogroups.com
>> > from the same email address that you subscribed.
>> > Text in the Subject and Body of the message will be
>> > ignored.
>> >
>> >
>> > Your use of Yahoo! Groups is subject to
>> > http://docs.yahoo.com/info/terms/
>> >
>> >
>>
>>
>> __________________________________
>> Do you Yahoo!?
>> SBC Yahoo! DSL - Now only $29.95 per month!
>> http://sbc.yahoo.com
>
>
>To UNSUBSCRIBE, just send mail to:
> basicstamps-unsubscribe@yahoogroups.com
>from the same email address that you subscribed. Text in the Subject and
Body of the message will be ignored.
>
>
>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
>
> I want to beable to speak into a mic with words such as "turn
> on", "move left", etc... Im sure its not going to be easy but thats
> what i am going to do but i need to know how first??
Nobody can tell you how to build this because, in order to know how,
they would have to do it themselves first!
>what would you recomend that i get to do this?
An off the shelf unit. There have been a ton of great suggestions for
off the shelf items that may do what you want.
> and how whould i sample the data to find a match?
You need an A/D sampling the electrical signal from your mic. But as
Dennis pointed out it needs to be a high speed A/D. For the simple
fact that you want it to recognize a complex phrase like "move left",
a Stamp won't have enough capability. You will need another platform.
This is a bit presumptuous- but the questions you are asking lead me
to think you don't have a lot of experience with projects like this.
If that is the case, a voice recognition system is not the place to
begin learning about this. Start small- I know it's a silly product,
but if you can build "The Clapper" you would have a good start on what
it would take to make a voice recognition system.
Neil
looks good to use, and im sure ill be able to manage with that.
http://www.imagesco.com/catalog/hm2007/SpeechRecognitionKit.html
. i might also be using PIC or if i dont get it, ill play around with
the basic stamp. if you have any more suggestions please tell me.
thanks =-)