matthiasm@0: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
matthiasm@0: % Audio Degradation Toolbox
matthiasm@0: %
matthiasm@0: % Centre for Digital Music, Queen Mary University of London.
matthiasm@0: % This file copyright 2013 Sebastian Ewert, Matthias Mauch and QMUL.
SebastianEwert@4: %
matthiasm@0: % This program is free software; you can redistribute it and/or
matthiasm@0: % modify it under the terms of the GNU General Public License as
matthiasm@0: % published by the Free Software Foundation; either version 2 of the
matthiasm@0: % License, or (at your option) any later version.  See the file
matthiasm@0: % COPYING included with this distribution for more information.
matthiasm@0: %
matthiasm@0: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
matthiasm@0: 
matthiasm@0: % Note: Some degradations impose a delay/temporal distortion on the input
matthiasm@0: % data. The last example shows, given time positions before the
matthiasm@0: % degradation, how the corresponding time positions after the degradation
matthiasm@0: % can be retrieved
matthiasm@0: 
matthiasm@0: %%
matthiasm@0: 
matthiasm@0: clear
SebastianEwert@4: close all
matthiasm@0: 
matthiasm@23: addpath(genpath(fullfile(pwd,'AudioDegradationToolbox')));
matthiasm@0: 
matthiasm@0: pathOutputDemo = 'demoOutput/';
matthiasm@0: if ~exist(pathOutputDemo,'dir'), mkdir(pathOutputDemo); end
matthiasm@0: 
SebastianEwert@4: filenames = {
SebastianEwert@4:     'testdata/RWC_G39.wav';
SebastianEwert@4:     'testdata/RWC_G72.wav';
SebastianEwert@4:     'testdata/RWC_G84.wav';
SebastianEwert@4:     'testdata/RWC_P009m_drum.wav';
SebastianEwert@4:     'testdata/RWC-C08.wav';
matthiasm@7:     'testdata/session5-faure_elegie2c-001-0.wav';
matthiasm@7:     'testdata/175234__kenders2000__nonsense-sentence.wav';
SebastianEwert@4:     };
matthiasm@0: 
matthiasm@9: createSpectrograms = 0;
matthiasm@0: 
matthiasm@0: %%
SebastianEwert@4: % just copying original files to the demo folder
SebastianEwert@5: maxValueRangeVis = zeros(length(filenames));
SebastianEwert@4: for k=1:length(filenames)
SebastianEwert@4:     copyfile(filenames{k}, fullfile(pathOutputDemo,sprintf('00_Original_file%d.wav',k)))
SebastianEwert@4:     if createSpectrograms
SebastianEwert@4:         [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@4:         [s,f,t] = spectrogram(f_audio,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@4:         figure; imagesc(t,f,log10(abs(s)+1)); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('00_Original_file%d.png',k)))
SebastianEwert@5:         maxValueRangeVis(k) = max(max(log10(abs(s)+1)));
SebastianEwert@4:     end
SebastianEwert@4: end
matthiasm@0: 
matthiasm@0: %%
SebastianEwert@4: for k=1:length(filenames)
SebastianEwert@4:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@4:     
SebastianEwert@4:     % with default settings:
SebastianEwert@4:     %f_audio_out = degradationUnit_addNoise(f_audio, samplingFreq);
SebastianEwert@4:     
SebastianEwert@4:     % adjusting some parameters:
SebastianEwert@4:     parameter.snrRatio = 10; % in dB
SebastianEwert@4:     parameter.noiseColor = 'pink';  % convenient access to several noise types
SebastianEwert@4:     f_audio_out = degradationUnit_addNoise(f_audio, samplingFreq, [], parameter);
SebastianEwert@4:     
SebastianEwert@4:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_01_addNoise_file%d.wav',k)));
SebastianEwert@4:     if createSpectrograms
SebastianEwert@4:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@5:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_01_addNoise_file%d.png',k)))    
SebastianEwert@4:     end
SebastianEwert@4: end
matthiasm@0: 
matthiasm@0: %%
matthiasm@0: % Some degradations delay the input signal. If some timepositions are given
matthiasm@0: % via timepositions_beforeDegr, the corresponding positions will be returned
SebastianEwert@4: % in timepositions_afterDegr.
matthiasm@0: timepositions_beforeDegr = [2, 3];
matthiasm@0: 
matthiasm@0: % Processing the time positions even works without processing the audio data (f_audio_out is empty afterwards)
SebastianEwert@4: parameter.loadInternalIR = 0;
SebastianEwert@4: parameter.impulseResponse = [1 -1 1 -1 1 -1 ] / 6;  % some impulse response
SebastianEwert@4: parameter.impulseResponseSampFreq = samplingFreq;
matthiasm@0: parameter.normalizeOutputAudio = 1;
matthiasm@0: [f_audio_out,timepositions_afterDegr] = degradationUnit_applyImpulseResponse([], [], timepositions_beforeDegr, parameter);
matthiasm@0: 
SebastianEwert@4: for k=1:length(filenames)
SebastianEwert@4:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@4:     
SebastianEwert@4:     % time positions and audio can also be processed at the same time:
SebastianEwert@4:     [f_audio_out,timepositions_afterDegr] = degradationUnit_applyImpulseResponse(f_audio, samplingFreq, timepositions_beforeDegr, parameter);
SebastianEwert@4:     fprintf('degradation_applyFirFilter: adjusting time positions\n');
SebastianEwert@4:     for m=1:length(timepositions_afterDegr) fprintf('%g -> %g\n',timepositions_beforeDegr(m),timepositions_afterDegr(m)); end
SebastianEwert@4:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_02_applyImpulseResponse_file%d.wav',k)));
SebastianEwert@4:     if createSpectrograms
SebastianEwert@4:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_02_applyImpulseResponse_file%d.png',k)))    
SebastianEwert@4:     end
SebastianEwert@4: end;
matthiasm@0: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.snrRatio = 10; % in dB
SebastianEwert@28:     parameter.loadInternalSound = 1;
SebastianEwert@28:     parameter.internalSound = 'PubEnvironment1';
SebastianEwert@28:     f_audio_out = degradationUnit_addSound(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_03_addSound_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_03_addSound_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.dsFrequency = 4000;
SebastianEwert@28:     f_audio_out = degradationUnit_applyAliasing(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_04_applyAliasing_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_04_applyAliasing_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.percentOfSamples = 10;  % signal is scaled so that n% of samples clip
SebastianEwert@28:     f_audio_out = degradationUnit_applyClippingAlternative(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_05_applyClipping_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_05_applyClipping_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.compressorSlope = 0.9;
SebastianEwert@28:     parameter.normalizeOutputAudio = 1;
SebastianEwert@28:     f_audio_out = degradationUnit_applyDynamicRangeCompression(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_06_applyDynamicRangeCompression_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_06_applyDynamicRangeCompression_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.nApplications = 5;
SebastianEwert@28:     f_audio_out = degradationUnit_applyHarmonicDistortion(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_07_applyHarmonicDistortion_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_07_applyHarmonicDistortion_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.LameOptions = '--preset cbr 32';
SebastianEwert@28:     f_audio_out = degradationUnit_applyMp3Compression(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_08_applyMp3Compression_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_08_applyMp3Compression_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.changeInPercent = +5;
SebastianEwert@28:     f_audio_out = degradationUnit_applySpeedup(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28: 
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_09_applySpeedup_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_09_applySpeedup_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.intensityOfChange = 3;
SebastianEwert@28:     parameter.frequencyOfChange = 0.5;
SebastianEwert@28:     f_audio_out = degradationUnit_applyWowResampling(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_10_applyWowResampling_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_10_applyWowResampling_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.stopFrequency = 1000;
SebastianEwert@28:     f_audio_out = degradationUnit_applyHighpassFilter(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_11_applyHighpassFilter_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_11_applyHighpassFilter_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     parameter.stopFrequency = 1000;
SebastianEwert@28:     f_audio_out = degradationUnit_applyLowpassFilter(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_12_applyLowpassFilter_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_12_applyLowpassFilter_file%d.png',k)))    
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: for k=1:length(filenames)
SebastianEwert@28:     [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:     
SebastianEwert@28:     % There are four ways to specify the destination mean spectrum: 1. by
SebastianEwert@28:     % loading example files provided with the toolbox, 2. by using specific
SebastianEwert@28:     % noise "color" profiles, 3. by providing the destination mean spectrum
SebastianEwert@28:     % using the parameter destMagFreqResp, 4. by providing audio data from
SebastianEwert@28:     % which the destination mean spectrum is computed.
SebastianEwert@28:     parameter.loadInternalMagFreqResp = 1; 
SebastianEwert@28:     parameter.internalMagFreqResp = 'Beethoven_Appasionata_Rwc';
SebastianEwert@28:     f_audio_out = degradationUnit_adaptiveEqualizer(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:     
SebastianEwert@28:     wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_13_adaptiveEqualizer_file%d.wav',k)));
SebastianEwert@28:     if createSpectrograms
SebastianEwert@28:         [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:         figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_13_adaptiveEqualizer_file%d.png',k)))
SebastianEwert@28:     end
SebastianEwert@28: end;
SebastianEwert@28: 
SebastianEwert@28: %%
SebastianEwert@28: if 1
SebastianEwert@28:     for k=1:length(filenames)
SebastianEwert@28:         [f_audio,samplingFreq]=wavread(filenames{k});
SebastianEwert@28:        
SebastianEwert@28:         [parameter.audioDataForDestMfcc,parameter.audioDataForDestMfcc_sf]=wavread('testdata/RWC_P009m_drum.wav');
SebastianEwert@28: 
SebastianEwert@28:         % After this unit, the mean MFCC vector of f_audio_out is almost identical to
SebastianEwert@28:         % the one of RWC_P009m_drum.wav
SebastianEwert@28:         parameter.visualizations = createSpectrograms;
SebastianEwert@28:         f_audio_out = degradationUnit_applyMfccMeanAdaption(f_audio, samplingFreq, [], parameter);
SebastianEwert@28:               
SebastianEwert@28:         wavwrite(f_audio_out,samplingFreq,16,fullfile(pathOutputDemo,sprintf('Unit_14_applyMfccMeanAdaption_file%d.wav',k)));
SebastianEwert@28:         if createSpectrograms
SebastianEwert@28:             [s,f,t] = spectrogram(f_audio_out,hamming(round(samplingFreq*0.093)),round(samplingFreq*0.093/2),[],samplingFreq);
SebastianEwert@28:             figure; imagesc(t,f,log10(abs(s)+1),[0 maxValueRangeVis(k)]); axis xy; colormap(hot); ylim([0,8000]); colorbar; print('-dpng', fullfile(pathOutputDemo,sprintf('Unit_14_applyMfccMeanAdaption_file%d.png',k)))
SebastianEwert@28:         end
SebastianEwert@28:     end
SebastianEwert@28: end
SebastianEwert@28: 
SebastianEwert@28: 
SebastianEwert@28: