voice recognition

jav2cube

Hi,
I'm new here. There is anyone or any topic that refer to doing a voice recognition module for basic stamp??
I hope you can answer me
Thanks

Mike Green

There was a brief discussion a few weeks ago on this topic. Basically, the Stamp doesn't have the processing power to do it. There were some simple voice recognition programs for the Apple II and similar 6502 processors years ago. They used a simple zero-crossing detector to look at the dominant frequencies in the voice input, did some normalizations (to make words the same length), then did a table lookup. You would teach them by repetitively saying the same word, then doing the same for other words. Usually, they'd work ok for 10 or 15 words like the digits and a couple of distinct commands, but not with great reliability. Fun to experiment with. Anyway, the 6502 used a few thousand bytes of memory for this process and the routines were tightly coded in assembly language. The Stamp just doesn't have the memory or the speed to do this.

Fe2o3Fish

It won't be "cheap" but you can check out the VR Stamp from Sensory, Inc.
www.sensoryinc.com/html/products/vrstamp.html

On the other hand, I remember that RadioShack sold a voice recognition chip
that did speaker independent (I think) recognition for the 10 digits and a few
other key words. Alas, that was some time ago and I don't remember if it
worked well or not. I sort of doubt it or it filled just too narrow of a niche 'cause
R/S didn't sell the chip for but a year, maybe two. (Yes, I know they sold a
voice synthesis chipset but this was voice recognition on a small DIP chip.
Probably just a zero-crossing counter style recognition -- hard to say)

True, the Basic Stamp can't do VR but since you're asking for a VR module you
are recognizing that too. Still, the Basic Stamp can surely control a VR module.

You'd be surprised what Googling the words "voice recognition chip" comes up with.
Google is your friend!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
-Rusty-
--
Rusty Haddock = KD4WLZ = rusty@fe2o3.lonestar.org
**Out yonder in the Van Alstyne (TX) Metropolitan Area**
Microsoft is to software what McDonalds is to gourmet cooking

Post Edited (Fe2o3Fish) : 8/23/2006 5:33:17 AM GMT

Chris Savage

Fe2o3Fish said...(trimmed)
On the other hand, I remember that RadioShack sold a voice recognition chip
that did speaker independent (I think) recognition for the 10 digits and a few
other key words. Alas, that was some time ago and I don't remember if it
worked well or not. I sort of doubt it or it filled just too narrow of a niche 'cause
R/S didn't sell the chip for but a year, maybe two.

Rusty,

·· I actually purchased that chip (which was around for about two years) back in the late 80's early 90's.· It was Speaker-Independant and the commands were more geared toward robot control, since it recognized words like "Forward", "Left", Right", etc.· I don't think it worked very well for me, but that was a long time ago.· It may be possible to find it in a surplus house (such as B.G. Micro or Electronics Goldmine), but I don't recall the part number.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Chris Savage
Parallax Tech Support
csavage@parallax.com

jav2cube

Thanks Everyone that reply me... Ireally appreciate that.

I found something interesting, is not wht i was looking for but it makes me fell close to it.

This is a PPS which is about a alarm clock with voice recognition, and in one method it uses the VR voice recognition module with the BASIC STAMP.

I hope you can teel me what do you think about this.

Once again.. THANKS

Kirk Fraser

One alternative is to buy a "clapper" to supply simple audio control to your Stamp.· Not certian on availability but "Dragon Dictate" used to supply chips for Apple's.

Proper voice recognition (VR) might be doable on a modern 1GHz or better PC but the available VR software such as "Dragon Dictate" and built-ins to Microsoft Office and such aren't it, at least not with my voice. H-P has a VR program answering its phones for printer support and it can't even recognize my "YES!" Some companies may risk sales on VR but it's not really ready for speaker-independant recognition yet.

The way to do it is:
1) Frequency analysis using an Fast Fourier Transform, Hartley, or Wavelet analyser.
2) Table look-up of phonetic patterns which are not of fixed size thus requiring guided learning (far more work than just feeding a Neural Net).
3) Higher level phoneme to word lookup and grammar based spelling checker.
4) Presentation software to insert the result where you want it or control what you want.

Not least in your thoughts should be the warnings in the Bible about the image that speaks, the abomination of desolation, the image of the beast, and the components of electronics listed in Daniel.

Kirk Fraser

Post Edited (Kirk Fraser) : 8/23/2006 5:25:06 PM GMT