Voice Command Recognition System

[hitusharpatil]

Hello everybody,
I am Tushar doing project "Voice Command Recognition System" using Atmega32 for switching Electrical devices. basic aim of my project is to design a voice command recognition system and its application will be switching electrical devices. Suppose a person is driving a car and he want to switch on AC/light/music system then he will give commands like 'light on'/'AC on'/'AC off' and system will work accordingly.

I devided this project in following steps.

1) Take input from user using mic in analog form

2) Convert analog signal to digital signal by ADC (Analog to Digital Converter)

3) Match digital signal with memory content ( For this part I'll store different sample commands commands in memory and for matching I may use Eucledian distance formula)

4) Give output of microcontroller to Relay Switch

My main problem is that I dont know how to store Sample voice commands (Fingerprints) in Microcontrollers Memory so that they can be compared with incomine voice Signal. Please Help me.

Can you send me some sample Code?

 

[Oznog]

Well, Microchip has a speech recognition library you can look at. It's actually a product and as such there's no source assembly. However, there is a broad description of how it works.

IIRC keywords are not stored in a literal format; it's not a sound file anymore but reduced to some sort of phonetic fingerprint. This makes the data considerably smaller and it's preprocessing the steps that would otherwise need to be performed at runtime on the keyword(s) we need to compare against anyways.

As you may guess recognition is actually quite tricky, if I try to match the oscilloscope waveform for "hello" with what I spoke earlier they're quite difficult to match up really, I wouldn't be able to visibly tell you the word from looking at the waveform. Really the code needs to understand the content of the sound.

 

[donniedj]

And enters DSPs. These have the speed and power to do and are already been doing voice recognition. Many cell phones and new cars already do this. If you are trying to market a new concept, get in line. But if its just a project, then you can try using something like a VoiceDirect 364 module from Sensory. Then you can still develop the code for user interaction.

Is your goal to control equipment by voice or voice recognition? Dumping data into an EEPROM is trivial. How can this be your main problem?

 

[xcuseme123]

hi... guys...
am a elecronics student....
i need some seminar topics on microcontroller,satellite communication,..etc....
here is my id>> xcuseme123@gmail.com<<<.....
if u lik to share contact plz.....

 

[hitusharpatil]

And enters DSPs. These have the speed and power to do and are already been doing voice recognition. Many cell phones and new cars already do this. If you are trying to market a new concept, get in line. But if its just a project, then you can try using something like a VoiceDirect 364 module from Sensory. Then you can still develop the code for user interaction.

Is your goal to control equipment by voice or voice recognition? Dumping data into an EEPROM is trivial. How can this be your main problem?

I want to do this as my Final year Project and my goal is to control Equipments by voice Commands.
I want to store Voice Fingerprints into memory so that it can be compared later with incoming voice samples. Can you please give me some sample code to dump voice Fingerprints into memory?

 

[prash_bhatia]

hey even i am trying something based on this

see wt u ppl have thot is
saving data for a particular specific use only
for eg "open" and "close"
so error or ur considerations are less

our audo range is from 20 hz to 20 khz

and words have specific freq as per use of stress words and consonants
so accordingly save freq values of the two words "open" and "close" to ur memory location as ur reference and compare given command freq with this

accordingly go ahead once command has been recognised


have nt tried this practically

so pls assure it before actually applying

luking forward for more ideas

thank u

 

[blueroomelectronics]

Please don't use txtspeak.