mini voice recognation project

[aliraza786]

hello i have a wage idea of mini voice recognition system.. i have chip which produce 4 different sounds. is that possible that i convert this sound to digital data using ADC0808 and save it in micro controller. when i give input such a signal i compared it with the stored one t if match its ok if not nothing will happen.... is that possible practically .? or any more suggestion is most welcomed... thank you in advance
m familiar with 89c51/52/55 controllers.

 

[Gofs]

It shouldn't be difficult. The question is about the comparison algorithm. If there will be only four the same sounds and generally their graph in the time domain won't change then comparision can be done quite fast on simple microcontroller. The worst thing in this kind of solution will be the noise. If the signal will be clear each time you want to detect it then if will be ok. If there will be additional noise then you will need probably at least some filter. The filter contruction and later the detection algorithm will depent on the complexity of the signal and the noise level. You could pass some more information about the signal, its FFT and sampling rate you would like to use.

 

[aliraza786]

thnkyou so much gofs. now what i understand is this that i store array of input signal in the for of decimal data in microcontroler. then again take input and comapre it with the saved one. but the problem is that how can i know when the input comes. mean i will constantnly looking at some port for data so how can i determine when the ecxact signal is coming.

 

[tpetar]

The circuit has been designed to develop a speech filter that will improve the signal processing circuit for optimizing speech recognition.

**broken link removed**

**broken link removed**

 

[aliraza786]

https://www.google.com.pk/imgres?um...w=200&start=0&ndsp=19&ved=1t:429,r:0,s:0,i:69

this the circuit which m using to generate 4 sounds. does this circuit output still need a speech filter..?

 

[Gofs]

I didn't know that this is speech? As far as I know simplest speech recognition is related with calculation time between points when signal passes zero level. I have never done this but on my university few yars ago some people were developing this method. And of course human speech is withing given fraquency range which is narrower than frequency range of sound heard by the human so you can use filter and make the sampling frequency lower.

Please pass more information about this sound characteristics.

 

[tpetar]

Thread is named "mini voice recognation project".

this the circuit which m using to generate 4 sounds. does this circuit output still need a speech filter..?

Please describe sounds in memory.

 

[Gofs]

Thread is named "mini voice recognation project".

Sorry, I don't know how it escaped my attention.

Generally you don't need a speech filter but then you must to have faster sampling rate. And later implement more complicated algorithm and you need more CPU power. You can combine this filter with microphone's amplifier which you will probably use.

About detection of voice in time. Is it possible to implement 'a button' which when it is pressed the device will start sampling sound signal? Or it must to be detected all the time?

If it is voice forget about simple comparision like I mentioned in my first post. This could be done only when sound from source is almost perfectly the same as sampled one stored in memory.

 

[aliraza786]

GOFs thnkyou so much.... it is sound which i have saved in computer and exactly the same sound will come and i have to make sample comparison. and button cant be used for this ...it must be detected all the time....... how can i do sample comparison in matlab.?

 

[Gofs]

Unfortunatelly I don't have experience in Matlab but I suppose yes.

 

[Kurenai_ryu]

well, recognize a simple frequency could be made with any microcontroller, by implementing a frequency counter.

but recognize more advanced sounds, like speech is a more difficult task,

this falls in the DSP (Digital Signal Processing) Area, which requires more complex and faster microcontrollers (like ARM or PIC32 or ARM32, not a small 8-bit microcontroller) or a PC directly (the easiest way to do DSP in a PC is with MATLAB i think...).

in DSP a common way is to take the voice signal to frequency (like a fourier transform or FFT) and analyze the peaks and valleys of frequencies to determine some parameters you want to recognize (like pitch and harmonics) but duration could be a problem...

so please determine your exact needs to begin your work.