need some help with this project

[bcktang]

the diagram in this link
**broken link removed**

is 24volts too high as the input at VDC???

i m using a microcontroller to control the motor to move left and right, as well as faster, the input for the microcontroller is the voice

"left", "right", "faster" and "slower" will b the input.

this is what i thought of,
ask a few friends to say the words mentioned above to a mic,
which goes through ADC, through ADC, i get the binary values, change to hex n store them into several memory location,

during operation, when a person says the word above, through the mic, it goes to ADC, then using programmes, it is compared to the stored values,

i think this will work, any comments?

BTWneed help, how to obtain the frequency of a certain word ? HELP PLS,

thanks

 

[Optikon]

Your idea sounds good in theory but you will have LOTS of problems getting that to work. Some of the reasons are:

You didnt mention the sampling rate of your ADC (It matters!)
Assuming you are sampling sufficiently fast, how many samples will you store?

1) Someone speaks and what is captured is several samples of their speech + noise.

2) When that person does it again, their inflection might be slightly off so, what your hardware will read is almost the same speech + noise again which is different than the original noise captured.

What if they speak the command 1 millisecond faster than when they did when you stored it? How will you handle the compressed version?

This is why voice recognition is so difficult to achieve.
You will need to do some heavy filtering and you will need some smart algorithms to compare stored values against input values to decide if they are "close enough" to allow it to execute as a command.

You may find that you need to store so many speech samples that the whole thing becomes pretty impractical unless you are willing to turn this into a full-blown voice recognition project?

I seem to recall that there are some IC's out there that you can wire up to this kind of thing and it will do some crude voice recognition and give you outputs. You might want to consider that.

Just my 0.02

 

[bmcculla]

Digikey has a voice recognition development Kit for 129$ VEKIT-ND that will make speech recognition easier. I wouldn't bother trying to make your own speech recognition algorithms they're extreamly complicated.

If you are looking for something related but a little easier you could have the microcontroller recognise the tone of a note that a person sings: High vs Low notes. By using an array of bandpass filters (allows waves within a certain frequency range to get through), rectifiers (a diode to cut off the negative part of the waveform) and integrators (turns the rectified ac into a DC level related to the AC amplitude), one set of these for each frequency band you want to recognise, you can identify the dominant frequency of a voice. All this could be done in digital filters too if your micro has the processing power.

Hope this helps

Brent

 

[john1]

Maybe simpler noises would be easier to recognise ...

SSSSS ... for stop
G ....... for go
RRRRR ... for right
FFFFF ... for left

or something like that?