Speech Recognition control

Status
Not open for further replies.

vincent_85

New Member
Hi everyone,
I'm currently doing a project on speech recognition to control various electronic/electrical appliances(might include robotics stuff also). I've known of the HM2007 speech IC but i intend to use other IC/microcontroller( Atmel, PIC) to do it.

Can anyone recommend me a low-cost and suitable microcontroller?actually i have intended to use the PIC microcontroller as i have knowledge on its programming.

Also, my circuit design is still currently under construction. Can anyone who has done speech recognition before or have knowledge on it give me tips/reference schematics/circuit design on building a speech recognition circuit?
where can i find resources on circuit design?
 

Use the HM2007 - you're not going to do it with a micro-controller!. But bear in mind it's probably not very reliable?.
 
Alternative?

how about other IC other than HM2007?any other speech recog. IC i can get easily?i'm currently living in Malaysia.
I've contacted Sensory Inc on their IC's(RSC-464) and dats the best i can find at the moment. Let's say i opt for this IC (RSC-464), is it necessary for me to use their development toolkit to do the programming and stuff?

I've also heard about the DSP freescale. Do u think i can come up with something with it?another problem is the circuit design
 

It took the AI types years to do a fair job of speech recognition. It is a very complex and difficult task. You are not going to write the software to do it. You are not going to whip up something on a microcontroller.

Look to the dedicated chips.

You also have to qualify what you need.
Is speaker dependent OK or do you want to do speaker independent. Is single work recognition OK or do you need to understand continuous speech.


 
i will start with speaker-dependent word recognition, unless the i can get my hands on IC dealing with speaker-independent.

how about dsPIC?is it possible?
if i really can't get hold of the HM2007?
any good IC's(maybe cheaper ones) i can use?
 
vincent_85 said:
i will start with speaker-dependent word recognition, unless the i can get my hands on IC dealing with speaker-independent.

how about dsPIC?is it possible?
**broken link removed**

I do not know anything about it.


An eval copy is $5 US. You can not use it in a product without a $2,500 license.

dsPIC30F5011 dsPIC30F5013 dsPIC30F6012 dsPIC30F6014
The dsPIC30's are $7 to $20+ US depending on your sources.

You could have goggled this without my help?

Anyway it does sound fun.

vincent_85 said:
if i really can't get hold of the HM2007?
any good IC's(maybe cheaper ones) i can use?
Attempting to be go cheap and ending up with a good system may not be possible.

EDIT: It seems that the dsPIC needs some help. The system also requires a Silicon Labs Si-3000 which is a codec (and other audio support stuff) chip. The good news is that it is under $2. The pfd for the speach software has the schematic for additional memory. It is starting to look like this is more complex then you would like.

In regards to what is inexpensive. This link lists parts that were around about 10 years ago. Perhaps some of them may be inexpensive.
 
Last edited:
The 33F is MUCH better than the 30F series. It is a different silicon process, 3.3v only, but much faster, lower power, and more advanced peripherals like a DMA controller. It's the same DSP core though.

I know the original version of the Speech Recognition Library was pretty badly flawed in the way it was unnecessarily hardcoded to work with only one configuration with the Si3000 and did the entire processing-intensive speech recognition process inside the hardcoded ISR. I did talk to them about that, I don't think they've reworked it yet though but they did see the problem.
 
BTW they don't charge $5 for the eval copy unless they mail it to you. You can download it, but you have to register first.
 
Oznog said:
BTW they don't charge $5 for the eval copy unless they mail it to you. You can download it, but you have to register first.

Thanks for the info. Too bad the did not do a better job on the software.

vincent_85: I think you best bet on getting it working is to use one of the ready to go chips like the HM2007. Anything else is going to be difficult at best.
 
Pic

My advisor has rejected me from using dedicated chips. So i have to think of a way to use the PIC/dsPIC or any other circuit that can recognize speech. Basically, i have to start from scratch.
This would be quite a headache.
If anyone has ideas,kindly tell me. Sorry for bothering.
 
Unless you are grad student in AI no one in his right mind should require that you write the software without using a speech lib or speech software toolset.

Try using the 30F series and the Microchip software we talked about earlier. He should accept that.
 
You could use some open source software like CMU Sphinx but you'll need to mess around a bit to get it to work.
**broken link removed**
 
Status
Not open for further replies.
Cookies are required to use this site. You must accept them to continue using the site. Learn more…