speech controlled leds

Thread Starter

Eric007

Joined Aug 5, 2011
1,158
Hi All,

i would like to learn about ASR (Automatic Speech Recognition) and be able to apply to a design...

For starters, i'm thinking of lighting 3 leds with 3 different words 'one', 'two', and 'three' for led1, led2 and led3 respectively!

so i'll have a voice input, a processing uC, an output (leds).

So i would like to know which PIC uC would be suitable for this kind of application? (also that PIC uC should have enough memory to store those words and a very good speed processing)

Also i would like you to provide me with some good ressouces about "voice recognition" so i learn about the kind of hardware i will need, ...different types of implementation,...

Thanks all for your comments!!
 

T.Jackson

Joined Nov 22, 2011
328
I built a voice recognition project from a kit many moons ago. It did exactly what you are describing; lit up a particular led for (n) word.

It didn't work so well though.
 

Thread Starter

Eric007

Joined Aug 5, 2011
1,158
ok jackson...i need to make this work so i can apply it to a much more complexed design...so i need to read up on it...
 

Thread Starter

Eric007

Joined Aug 5, 2011
1,158
lol...simple to do...that good to hear but can you give some hints, and ressources?...
 
Last edited:

MrChips

Joined Oct 2, 2009
30,714
Someone else had the same question. My answer was that IBM spent millions of dollars and may years working on this problem. Good luck!
 

kubeek

Joined Sep 20, 2005
5,794
Try looking for some open source speech recognition and see how they do it, and how demanding those libraries are so you can tell what cpu you would need.

Also you need to decide if you want general voice recognition, so that anyone saying the word will activate it, or you can do with sampling the users voice and then finding the best match, which is way simpler to do.
 

T.Jackson

Joined Nov 22, 2011
328
Someone else had the same question. My answer was that IBM spent millions of dollars and may years working on this problem. Good luck!
Reminds me of the story about a couple of scientists back in the 70's who spent a lot of time and money on a natural languages project.

They made a fool out of themselves. They expected to have an intelligent conversation with a computer!

This is not possible.

Worse still is we now have a division of people who feel that the trick is to teach AI / build upon its base with new information. The problem is that natural languages such as English use words with a abstract meaning, usually defined by their context. Pronouns are likely the biggest hurdle though. I say I, but am I implying you?

Something that only a human would ultimately pick up on.
 
Last edited:

justtrying

Joined Mar 9, 2011
439
I was considering a similar project... There are basic ICs limited to 40 words (ex/ HM 2007). Most commercial products are software based. Quality of recognition is always an issue. Have to be calibrated for individual voices. Accents are always a problem ;(
 

justtrying

Joined Mar 9, 2011
439
Reminds me of the story about a couple of scientists back in the 70's who spent a lot of time and money on a natural languages project.

They made a fool out of themselves. They expected to have an intelligent conversation with a computer!

This is not possible.

Worse still is we now have a division of people who feel that the trick is to teach AI / build upon its base with new information. The problem is that natural languages such as English use words with a abstract meaning, usually defined by their context. Pronouns are likely the biggest hurdle though. I say I, but am I implying you?

Something that only a human would ultimately pick up on.
so once they built cognitive computer I and I will know how to deal with it ;) (props)
 
Top