MFCC extraction of an audio file

17 vues (au cours des 30 derniers jours)
Mudit Batra
Mudit Batra le 30 Mar 2020
MFCC
How can we extract mfcc feature of an audio file (total 39 comprised of normal-delta-double delta) having window lenght of 25ms and window shift of 10ms?

Réponses (1)

jibrahim
jibrahim le 31 Mar 2020
Mudit,
Check out the mfcc function in Audio Toolbox.
[audioIn,fs] = audioread('Counting-16-44p1-mono-15secs.wav');
winLength = round(0.025 * fs);
overlapLength = round(0.015 * fs);
[coeffs,delta,deltadelta] = mfcc(audioIn,fs,'WindowLength',winLength,...
'OverlapLength',overlapLength,...
'NumCoeffs',13','LogEnergy','Ignore');
  2 commentaires
Mary Jeethu A J
Mary Jeethu A J le 26 Août 2021
Modifié(e) : Mary Jeethu A J le 26 Août 2021
Using this code I'd computed mfcc.And obtained 498*13 featured vector for cepstral coefficiets ,delta and delta -delta.13 is the number that represents the coefficients.Could you please let me know what 498 indicates?
I'd used an audio of 5 sec length. Is that number related to that? Or is it related to sometig like frames?
MathWorks Audio Toolbox Team
MathWorks Audio Toolbox Team le 26 Août 2021
Hi Mary,
Yes, 498 is the number of windows (frames) over which you computed the mfcc. It depends on the duration of your signal, and the window parameters specified on the function (Window and OverlapLength). For the same window length, the longer your signal the larger this number.

Connectez-vous pour commenter.

Catégories

En savoir plus sur Feature Extraction dans Help Center et File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by