mfcc - compute mel-scaled cepstral coefficients


mfcc (-i item) (-n ncoeff) (-l low_freq) (-h high_freq) (-r frame_rate) (-e|-F) (-p) (-d1|-d2) sfsfile


mfcc designs a mel-scaled filterbank for the supplied number of channels and processes the input speech signal through the filterbank into a set of mel-scaled cepstral coefficients. The method is based on that described in Deller, et al, Discrete Time Processing of Speech Signals. It uses a Hamming window and a forward FFT and and a specialised inverse DFT.

Options and their meanings are:

-I Identify program and version number.

-i item Select input item number.

-n ncoeff Select number of coeffients to calculate. Default: 16.

-l low_frequency Specify lowest frequency of filterbank. Default: 0Hz.

-h high_frequency Specify highest frequency of filterbank. Default: half input sample rate.

-r frame_rate Specify frame rate for output. Default 100 frames/sec.

-e Append the log signal energy to the output frame.

-F Store filter output rather than cepstral coefficents;

-p Do not pre-emphasise signal.

-d1 Add first order delta cepstral coeffs to output vector.

-d2 Add first and second order delta cepstral coeffs to output vector.


1.xx Any speech waveform.


11 Mel-scaled cepstral coefficients.


1.2 - Mark Huckvale
Fri Jul 09 14:54:38 2004