Audio data loading and splitting

How to load the audio dataset with subfolders in pwd, create ads and spilt them into train and testing set

[Walter Roberson]

I am not sure what you mean by "create ads"?

Answer 1

[Mathieu NOE]

hello

see example below - adapt to you own needs

%% Important Notice : 
% NB : wav files are not searched inside the main folder (yourpath) but only in subfolders (dirnames)
% in other words : this code works for wav files that are only located in
% subfolders , not in the main "yourpath" folder

% if S is empty that means no data file exist (matching the file filter in the searched subfolder)

%% define path
yourpath = pwd; % or your specific path
list=dir(yourpath);  %get info of files/subfolders in current directory
isfile=~[list.isdir]; %determine index of files vs folders
dirnames={list([list.isdir]).name}; % directories names (including . and ..)
dirnames=dirnames(~(strcmp('.',dirnames)|strcmp('..',dirnames))); % remove . and .. directories names from list

%% FFT options 

% if you are dealing with acoustics, you may wish to have A weighted
% spectrums / spectrograms
% option_w = 0 : linear spectrum (no weighting dB (L) )
% option_w = 1 : A weighted spectrum (dB (A) )
option_w = 0;     

% spectrogram dB scale
spectrogram_dB_scale = 80;  % the lowest displayed level is XX dB below the max level

NFFT = 512;    % fft buffer size
OVERLAP = 0.75;  % percentage of overlap (1 = 100%)


%% Loop on each folder
for ci = 1:length(dirnames) % 
    
    fileDir = char(dirnames(ci)); % current directory name
    S = dir(fullfile(fileDir,'*.wav')); % get list of data files in directory according to name structure 'Sheeta*.xlsx' 
    S = natsortfiles(S); % sort file names into natural order (what matlab dir does not well) , see FEX : 
               %(https://fr.mathworks.com/matlabcentral/fileexchange/47434-natural-order-filename-sort)
               
    %% Loop inside folder
    for k = 1:length(S) % read data in specified sheet
        [signal,Fs] = audioread(fullfile(fileDir, S(k).name)); % or use a structure (S(k).data ) to store the full data structure
        dt = 1/Fs;
        [samples,channels] = size(signal);
        time = (0:samples-1)*dt; % time vector 
     
        title_str = [fileDir ' / ' S(k).name];
        title_str = strrep(title_str,'_',' '); % replace underscores with blanks for better rendering
        
        
        %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
        % spectrogram demo
        %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
        for ck = 1:channels
            [sg,fsg,tsg] = specgram(signal(:,ck),NFFT,Fs,hanning(NFFT),floor(NFFT*OVERLAP));  
            tsg = tsg + NFFT/(2*Fs); % small time correction so that a fft buffer is time referenced at the buffer middle point
            % FFT normalisation and conversion amplitude from linear to dB (peak)
            sg_dBpeak = 20*log10(abs(sg))+20*log10(2/length(fsg));     % NB : X=fft(x.*hanning(N))*4/N; % hanning only

            % apply A weigthing if needed
            if option_w == 1
                pondA_dB = pondA_function(fsg);
                sg_dBpeak = sg_dBpeak+(pondA_dB*ones(1,size(sg_dBpeak,2)));
                my_title = ('Spectrogram (dB (A))');
            else
                my_title = ('Spectrogram (dB (L))');
            end

            % saturation of the dB range :  the lowest displayed level is XX dB below the max level
            min_disp_dB = round(max(max(sg_dBpeak))) - spectrogram_dB_scale;
            sg_dBpeak(sg_dBpeak<min_disp_dB) = min_disp_dB;

            % plots spectrogram
            figure();
            ah1 = subplot(3,1,1);
            plot(time,signal(:,ck));
            xlim([min(time) max(time)]);                       % make sure time and spectrogram have same x axis
            title(title_str);        % time domain plot (one channel)
            xlabel('Time (s)');
            ylabel('Amplitude');
            
            ah2 = subplot(3,1,2:3); 
            imagesc(tsg,fsg,sg_dBpeak);colormap('jet');        % spectrogram plot (one channel)
            xlim([min(time) max(time)]);                       % make sure time and spectrogram have same x axis
            axis('xy');grid on
            hcb = colorbar('vert');
            set(get(hcb,'Title'),'String','dB')
            
            %# find current position [x,y,width,height]
            pos2 = get(ah2,'Position');
            pos1 = get(ah1,'Position');

            %# set width of second axes equal to first
            pos2(3) = pos1(3);
            set(ah2,'Position',pos2)
            df = fsg(2)-fsg(1); % freq resolution 
            title([my_title ' / Fs = ' num2str(Fs) ' Hz / Delta f = ' num2str(df,3) ' Hz / Channel : ' num2str(ck)]);
            xlabel('Time (s)');
            ylabel('Frequency (Hz)');

        end
                
    end
        
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% A weighting for acoustic FFT Analysis
function pondA_dB = pondA_function(f)
	% dB (A) weighting curve
	n = ((12200^2*f.^4)./((f.^2+20.6^2).*(f.^2+12200^2).*sqrt(f.^2+107.7^2).*sqrt(f.^2+737.9^2)));
	r = ((12200^2*1000.^4)./((1000.^2+20.6^2).*(1000.^2+12200^2).*sqrt(1000.^2+107.7^2).*sqrt(1000.^2+737.9^2))) * ones(size(f));
	pondA = n./r;
	pondA_dB = 20*log10(pondA(:));
end

[Mathieu NOE]

hello again

problem solved ?