Subject:
|
Re: RCX Programming Questions and Sensor Ideas
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Sun, 29 Aug 1999 02:29:16 GMT
|
Viewed:
|
941 times
|
| |
|
|
In lugnet.robotics, Laurentino Martins writes:
> I think it would be more interesting to synthesise human voice, that is to make the RCX speak! :-]
Maybe AS interesting..
I also have a TI Speak 'N' Math that I would like to rip the guts out of, too.
Has anyone hacked a Speak 'N' Whatever before? Is it worth it?
Amos
Turn my address umop episdn to shake out the spamcake.
> At 09:32 27-08-1999 Friday , abieler wrote:
> > I hope someone reads this reply to a reply to a very OLD message.
> > I've been lurking for almost a month and, during a search, I found
> > a thread I can expand.
> >
> > In lugnet.robotics, lego-robotics@crynwr.com (Peter Hesketh) writes:
> > > In article <255199E650C3D11194BF00805FA785365101AB@cs03mail.BestBuy.com>
> > > , Cwikla, Brian <Brian.Cwikla@BestBuy.com> writes
> >
> > [snip]
> >
> > > > A voice recognition module. The module has, say, 8 memory banks where there
> > > > can be 3 or 4 seconds of 8-bit audio. Then, the on board voice recognition
> > > > processors analyze the sound, match it to the proper bank and then send a
> > > > number value to the RCX input port that it is attached to. I know this
> > > > sounds really far fetched and probably really expensive, but I thought it
> > > > would be cool to verbally tell my RCX to turn right etc...
> > >
> > > Voice recognition is much more difficult than comparing digitised sound
> > > with an 8-bit audio file. Just try recording and comparing the binary
> > > of 2 wav files of the word "Left"
> >
> > It is more difficult than this to recognize speech, but it can be done.
> >
> > And, in a 14-pin Plastic DIP!
> >
> > I have an old (early 90's ??) chip from RadioShack that contains a
> > fully functional 5 word (go, back, stop, turn left and right turn)
> > or 2 word (on/off, up/down etc.) selectable voice recognition system.
> >
> > Does anyone know if this or a similar product is still in production?
> >
> > The way this chip works is:
> >
> > 1. It filters the speech until it is a sqare wave with a changing
> > pulse width and frequency, but a constant amplitude.
> >
> > For each of the recognizable words, this wave will be *very*
> > similar between any two speakers, also *totally* unrecognizable
> > as speech by a human.
> >
> > 2. It uses this wave as a sort of "audio barcode." A barcode (or magstripe)
> > uses the *ratios* between pulses, instead of their frequency, because
> > the card can get swept through the reader quickly or slowly.
> >
> > Since people all speak at different pitches, the frequencies will
> > be different. To compare, again, with a barcode, a person with a
> > high pitched voice would "swipe" quickly, whereby eg. Barry White
> > would swipe slow.
> >
> > 3. Just like a barcode reader, the chip then determines whether a word
> > was said, if it was valid and which one it was.
> >
> > 4. To interface with the controlled circuit, it lowers one of 5 lines,
> > plus an 'enable' signal, for each word.
> >
> > The way it outputs which command was spoken will be a major obstacle
> > to using it with the RCX/CyberMaster.
> >
> > I'm not sure about how the inputs on these work, but, maybe a resistor
> > ladder where each word has it's own unique resistance.
> >
> > Also, I think my chip requires a 'dual rail' (eg. +5v and -5v) power supply.
> >
> > Perhaps there is a better chip on the market that will be more compatible.
> >
> > -------------
> > Amos Bieler
> > Turn my address umop episdn to shake out the spamcake.
> >
> > > > Why not a sensor that relates to a remote beacon. The remote beacon is a
> > > > battery operated, stand alone unit that emits a IR pulse/signal every x
> > > > amount of time. The sensor connects to the RCX has four detectors on it all
> > > > facing in opposite directions. The sensor receives the pulse and the two
> > > > detectors that receive the strongest signal (from the direction the beacon
> > > > is) derive some number that is sent to the RCX. Multiple beacons would have
> > > > different 'carrier' signals so that the program can determine from what
> > > > beacon the signal came. I think the reason I thought this one up is for a
> > > > very specific device I was building and am not sure how useful this would be
> > > > in a general sense. However, I can't imagine this being a
> > > > complicated/expensive device to build.
> > >
> > > IR has problems with obstacles in a room. A much better method is
> > > ultrasonic beacons as they transmit round corners and under chairs. You
> > > would not use 4 directional sensors, but measure the time delay for a
> > > distance measurement.
> > >
> > > --
> > > Regards - Peter Hesketh, Mynyddbach, Mon.
> > > Forty reasons why a dog is better than a woman: number 6
> > > "Dogs don't expect you to call when you are running late."
> > --
> > Did you check the web site first?: http://www.crynwr.com/lego-robotics
>
>
> Laurentino Martins
>
> [ mailto:lau@mail.telepac.pt ]
> [ http://www.terravista.pt/Enseada/2808/ ]
>
> --
> Did you check the web site first?: http://www.crynwr.com/lego-robotics
|
|
Message is in Reply To:
17 Messages in This Thread:
  
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
