How do Mel Frequency Cepstrum Coefficients work?

Question

I allready have FFT and pitch + absolute frequency calculated in real-time from input of microphone. Now I want to calculate the timbre.

I saw Mel Frequency Cepstrum Coefficients - MFCCs but I didn't understand it very well. Can someone give me some tips on this..

Donia

Answer 1

MFCCs combine consideration of aspects of human hearing (logarithmic frequency perception, the mel scale) and physics of musical instruments (these systems often have well defined overtones that are harmonic -- which is why the MFCCs use the FFT of the FFT), to give a simplified representation of the timbre of an instrument (where the fundamental frequency and loudness are factored out).

One could write endless pages on this topic, and there are many available on the web, so a more specific question that explains clearly what you want to know would be helpful. The algorithm for calculating MFCCs is listed at the top of the wikipedia page.

Takahashi

tks for the answer. So if I want to obtain the timbre, the MFCC is the way to go right? – Psychokinesis

Yes, I think so, especially as a first attack on the problem. It's easy to implement, there's a lot written about it, and it's a standard against which other approaches are compared (and it usually does well in these comparisons), so it's a good choice. – Takahashi

In wiki the algorithm is: 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. I understand everything but step 2: using triangular overlapping windows What should I do, pass frequency to mel scale, but what is this triangular overlapping window? – Psychokinesis