Voice command recognition

Voice command recognition (also referred to as “automated voice command recognition”, “computer voice recognition”, or in a slightly altered form “voice recognition”) is the process of converting audio signals specific to the human voice into a sequence of words, using an algorithm implemented as a computer program. Next we will use the abbreviation VCR to refer to the current title.

In June 2006, I’ve developed such a computer software, for my master dissertation thesis.

In the first step it receives the human voice from a microphone. Next it applies a Fast Fourier Transformation on the data set recorded with the microphone. The result is that for every sound it records, it processes a sequence of numbers based on the distribution of the frequencies.
You can see this as a number based descriptor, for various audio signals.
These sequences can then be applied to a neural network for the purpose of recognizing them.

This is only the main idea, more work has been actually done to cut representative voice signals out of the continuous stream coming from the microphone. The processor cost is quite high in this phase and it needs improvement, but overall the algorithm works well.

Here is a short video with me showing this software:

Hope you like it

Radu Motisan

This article has 11 Comments

  1. Altought not quite understanding all that is going on, it’s allways nice to put a face (and a voice), on a name! ๐Ÿ˜€

  2. Thanks Vasco ๐Ÿ™‚

    Regarding the content, I might try to port it on Windows Mobile, I still need to see how the hardware can handle it, and to get some time for this, of course.

    Also there are some additional improvements required to transform a didactic application into a commercial one.

  3. by seiing the mobile its a latest content involved but it s still need to see the ore ppts and web based relating these projects

  4. Thanks for giving such information but we need more information because as we are doing a project in this so we require extra ppts an some well preferred matter considering 30to 40 pages

  5. Hi.
    Elsewhere you have mentioned the posibility of porting your software to ATMEL microcontrollers.
    I was wondering if you have had any success in this area?

    Great work by the way,

  6. Thanks Jim, I had some success but I’ve put the project on pause before completing it, because of missing time.

    I hope to be able to make a nice post about that soon.

  7. Nice work. But wouldn’t it be better to port your software on a DSP since there is a lot of calculating required (Fourier Transform). With this solution you would be able to put the main intelligence on a ยตC board.

Leave a Reply