speech recognition using microcontroller

[Emantra]

hi everybody,

I am doing a project on speech recognition using the microcontroller which then has to open or close the door.
but the problem that i am facing is that i am pretty new in this field.

so here is what my speech recognition ckt has to do
1. maintain a database of 2 or 3 person sound of two command OPEN and CLOSE.
2. when the authorized person speaks on the mic saying OPEN or CLOSE has to turn on the DC motor.
3. if an unauthorized person speaks to the mic then generate a buzz sound.

when these task were performed by the microcontroller then it has to dirive a DC motor in order to open or close the door. for this purpose i have thought to use the H-bridge.

and also i am posting a h-bridge figure for u guys to be more clear.

so can i get any help on this matter will be very great of u guys

thanks for any reply

 

[checkmate]

Speech recognition is NOT a trivial task. Furthermore, it's still an experimental field with varying levels of success and all sorts of fancy algorithms.
Standard microcontrollers are mostly incapable of anything involving speech or image/video processing. I would guess that at minimum, you would need a DSP. Of course, ready-made solutions are the best as described by others.

Added : I love Nepal, best place I've ever been.

 

[Emantra]

r u sure that a microcontroller cant perform a speech recognition task.

well if it is yes then can u give more idea on it, rather than a speech reconition chip coz i am not allowed to use these kind of chip, do u have any idea

n yes nepal is the best place to be

thanks for any reply

 

[checkmate]

Is this a commercial project or a school assignment? If it's a school assignment, go back to your lecturer and tell him it can't be done. IMO, any speech processing algorithm that is precise enough to distinguish the voices of different people is fit for a master's, or even phd thesis.

 

[Emantra]

well this is my final yr project. and now i cant change to any other topic to coz the 1st internal assesment for this project has been completed so i cannnot change it be any means so i have to stick to it anyway. so if u could find any kind of help will be appreciated


thanks

 

[Someone Electro]

I think microchip made speach recogniton libarys for dsPIC.But there not free

 

[Joel Rainville]

I would start by looking here : **broken link removed**

There are 2 speech recognition libraries sold by Microchip for its dsPIC30F series of DSPs, but you can't even get a free demo copy...

Freescale (Motorola) seems to offer a broader range of DSPs, and they claim to integrate a speech recognition peripheral into their chips : https://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=DSP56858&nodeId=0127956292

Hope that helps...

 

[evandude]

I seriously hope you have taken some DSP courses :lol:

this is definitely a project I would never even consider attempting if barred from using commercial speech recognition chips... even if you could use them it would be challenging enough!

 

[samcheetah]

i appreciate what you are planning to do. it isnt something that is impossible. but i cant say that it would be good for you or not. what are you doing right now? an engineering degree? and do you have the resources, like a DSP kit?

and yes a microcontroller cant do speech recognition. actually if you have studied DSP you would know that the main operation in signal processing is MAC (multiply accumulate) and and microcontrollers arent good at multiplication. you have the choice to use a DSP board or an FPGA board.

and have you done any speech recognition in MATLAB or some other simulation software? if you have done such a thing, only then will you be able to move on to mapping speech recognition algorithms on some hardware.

if you didnt understand the above discussion i would say that you leave the project. but remember its always a good time to start a thing. if you have enough time then you could learn all the things that are required provided that you have good instructors

i hope that helps

 

[Emantra]

i am doing an engineering degree rite now, well we do not have DSP kit.

one of my freind told that he can write a program in MATLAB for the simulation of speech recognition.

but i dont know wat can we do with that simulation. also my tutor for this project told me to write a program in C or VB for the speech recognition.

is this a good idea?



thanks

 

[samcheetah]

you do simulation to design a system. like for instance you would have studied about FIR filters. they are the simplest applications of digital signal processing. when designing an FIR filter you have to choose the right set of coefficients. if you simulate your filter in MATLAB you can see the response of your filter with different sets of coefficients. when you are satisfied with one set of coefficients, you map the filter into some hardware board.

but MATLAB is a simulation software. if you want to make a good looking project then you should use Visual C++. there are many libraries of VC++ that are related to speech recognition. go to microsoft's website and download the SAPI tool kit, its about 60MB big. read the documentation and start coding! have fun.

if you dont have a DSP kit in your university (college or whatever) then buying one might be a bit heavy on your pocket. so its better if you use a PC for the speech recognition. then you can interface your control circuit with the PC through parallel or serial port.

i think this would be a good project

 

[checkmate]

Yes. Since you don't have much choice, the best way is to drop the microcontroller idea and make use of the PC's powerful processor and plentiful RAM. You'll have more libraries to work with too. However, you may retain the microcontroller solely for the purpose of data acquisition and data transfer.

The main problem with microcontrollers in speech processing is not only their lack of multipliers, but more importantly, their limited RAM. Higher-end PICs give you 4kB of memory. But the human voice is commonly regarded as having a bandwidth of 4kHz. Assuming 8-bit ADCs are used, the RAM can only last about one second. Furthermore, although my knowledge with speech processing is limited (I do know basic DSP though), I would associate it more with transforms, correlators and markov processes, which need even more RAM.

 

[Emantra]

Well going through all ur suggestion I think that I should now change the concept of doing my project.

So I think that it is a better idea to write a code on C/C++ or the VB than to run after a microcontroller and simulating in MATLAB.

So do u guys know anylinks on this matter

Any help will be appreciated.

 

[GatorGuy]

There is a voice package that you can get for free for VB. I have been using it for some time now. It is pretty accurate and is easy to work with. It even comes with text to speech. I'm at my other office right now and I don’t have any links but I will try when I get back to my main office. You could still work with the microcontrollers to work on several doors and have one main server running it all.

 

[Emantra]

I would really appreciate that.

I didnt know that VB had provided such feature too.

I'll be waiting


thanks

 

[GatorGuy]

For starters:

**broken link removed**

 

[samcheetah]

i dont know if you are more comfortable with VC or VB. but if you want something for VC then here is the link for the SDK.

https://www.microsoft.com/speech/download/sdk51/

 

[Emantra]

Well thansk for the links guys.

I havent yet gone through these site coz I have to prepare for my internal assesment. So when my assesment finished I'd surely look into it.

Thanks for the reply.

 

[kinjalgp]

Here is what you are looking for. An 8051 like microcontroller with in-buit speech recognition functionality.
https://www.sensoryinc.com/

 

[samcheetah]

the problem with a completely embedded solution is the complexity. i dont say that this cant be done. but any embedded speech recognition system cant match the flexibility of using a PC. i mean everything is so easy if you do it with a PC. embedded systems are complex things. and at the undergraduate level nobody expects you to build multi-DSP systems or embedded speech recognition modules. you might run into many problems so its better if you use the PC.

my 2 cents 8)