Iterate speech frames through all the audio file in the folder

Hello, i have split audio file into frames and i want to iterate all the frames though all the audio file in the folder. After iterate, i want to apply feature like mfcc(calculate the mean) of all the audio file and put it in a matrix of rows and conlums(each row is mfcc mean for each file). Below is the code i m trying to make as what i need. Any help please.

clc;
clear all;
% Specify the folder where the files live.
myFolder = '/MATLAB Drive/AUDIO2/LONGhead/FC03_1head';
% Check to make sure that folder actually exists.  Warn user if it doesn't.
if ~isfolder(myFolder)
    errorMessage = sprintf('Error: The following folder does not exist:\n%s\nPlease specify a new folder.', myFolder);
    uiwait(warndlg(errorMessage));
    myFolder = uigetdir(); % Ask for a new one.
    if myFolder == 0
         % User clicked Cancel
         return;
    end
end
% Get a list of all files in the folder with the desired file name pattern.
filePattern = fullfile(myFolder, '**/*.wav'); % Change to whatever pattern you need.
theFiles = dir(filePattern);
AudioArray = cell(1, numel(theFiles));
i=1;
j=1;
o=1;
r=1;
u=1;
for k = 1 : numel(theFiles)   % NUMEL is more direct than LENGTH
    baseFileName  = theFiles(k).name;
    fullFileName  = fullfile(theFiles(k).folder, baseFileName);
    AudioArray{k} = audioread(fullFileName);
    [speech, fs] = audioread(fullFileName);
% do framing
    f_d = 0.03;
    f_size = round(f_d * fs);
    n = length(speech);
    n_f = floor(n/f_size);  %no. of frames
    temp = 0;
    for i = 1 : n_f
        frames = speech(temp + 1 : temp + f_size);
       %frames = X1(temp + 1 : temp + f_size);
        temp = temp + f_size;
        y=mfcc(frames, fs);
       % z(i, :)=max(fwht(frames))
        Z(i, j)=[mean(y)];
    end
end
    

Answer 1

[jibrahim]

Hi Camille,

The function mfcc breaks the signal into frames for you, so you do not need to perform the frame processing yourself. Also, consider using an audioDatastore, as it will probably simplify your file handling logic.

Here is an example:

% This datastore points to all your audio files
ads = audioDatastore(pwd,'Includesubfolders',true);

numFiles = numel(ads.Files);

frameLength = 0.03; % in seconds

numCoeffs = 13;
avrgValues = zeros(numFiles,numCoeffs);

for index=1:numFiles

    % Read an audio file
    [x,myInfo] =read(ads);

    % Take the first channel of audio here
    x = x(:,1);

    fs = myInfo.SampleRate;

    winLength = round(frameLength*fs);
    coeffs = mfcc(x,fs,"Window",hann(winLength,'periodic'),'NumCoeffs',13,'OverlapLength',0,'Logenergy','ignore');
    
    avrgValues(index,:) = mean(coeffs);

end

[Camille Dingam]

Thank you for your answer, it is really helpful. But beside, i want to extract LPC features too. Can i use frame and window within LPC too? Please show how to do the frame for LPC. Thank you again

[jibrahim]

If you are using a toolbox function for LPC computation, you should check the help of that function - it might have a window/frame parameter. If you are using your own code for LPC, then you might have to perform framing yourself and then performing your custom computations; it depends on your algorithm. The way you were performing framing looks good to me. You can also consider looking into the buffer function. It will break your signal into frames for you.

[Camille Dingam]

Thank you very much, it is really helpful