MFCC

Hi all,

long time reader, first time posting as a member.

My query relates to MFCC calculation. Generally speaking I have an ok
verbal understanding of MFCC's, however I never had a detailed
understanding.

Firstly, im trying to implement this in MATLAB (this being MFCC
calculation).

Im fairly all right at MATLAB, so the previous steps to MFCC calculation
such as windowing and framing are of no trouble to me.

Im not asking/looking for MFCC MATLAB code (although that does sound nice
if its well commented) as I have came across MFCC MATLAB code already on
the internet.  What I am looking for is a good step by step explanation of
how to generate MFCC's.  The end result would be that I understand the
explanation enough to code it myself in MATLAB.

Now I have googled MFCCs before, but I was wondering if anyone knows where
a good explanation exists (or if they have one themselves).

All going well, I'd be happy to post up my MFCC code once I've understood
and coded it.

I hope I have explained myself well,

Thank you for your time,

Mark

On Jul 30, 12:30 pm, "nerdlinger" <nerdlinger.gene...@gmail.com>
wrote:
> I hope I have explained myself well,

I think you should probably define the acronym "MFCC". It's not one
that is all that common.

> Im not asking/looking for MFCC MATLAB code (although that does sound nice
> if its well commented) as I have came across MFCC MATLAB code already on
> the internet.  What I am looking for is a good step by step explanation of
> how to generate MFCC's.  The end result would be that I understand the
> explanation enough to code it myself in MATLAB.

Assuming you mean "Mel-frequency cepstrum coefficient,", from
http://en.wikipedia.org/wiki/Mel_frequency_cepstral_coefficient:

---
MFCCs are commonly derived as follows:[1]

   1. Take the Fourier transform of (a windowed excerpt of) a signal.
   2. Map the powers of the spectrum obtained above onto the mel
scale, using triangular overlapping windows.
   3. Take the logs of the powers at each of the mel frequencies.
   4. Take the discrete cosine transform of the list of mel log
powers, as if it were a signal.
   5. The MFCCs are the amplitudes of the resulting spectrum.
---

Seems pretty straightforward.

Jason

> Im not asking/looking for MFCC MATLAB code (although that does sound 
> nice

Try here:
http://labrosa.ee.columbia.edu/matlab/rastamat/
or here
http://www.aquaphoenix.com/research/matlab/

Please note that they produce different results.

Hope that helps

Erich