function example_output = exp_dietz2011(varargin)
%EXP_DIETZ2011 Experiments from Dietz et al. 2011
%
% EXP_DIETZ2011(fig) reproduce Fig. no. fig from the Dietz et
% al. 2011 paper.
%
% *Note**: The input signals used in this routine are not identical to
% the ones used in the original paper.
%
% The following flags can be specified;
%
% 'plot' Plot the output of the experiment. This is the default.
%
% 'noplot' Don't plot, only return data.
%
% 'fig3' Reproduce Fig. 3 panels a + b.
%
% 'fig4' Reproduce Fig. 4.
%
% 'fig5' Reproduce Fig. 5.
%
% 'fig6' Reproduce Fig. 6.
%
% 'fig6_spille2013' Reproduce Fig. 6 from Spille et al. (2013).
%
% Examples:
% ---------
%
% To display Fig. 3 use :
%
% exp_dietz2011('fig3');
%
% To display Fig. 4 use :
%
% exp_dietz2011('fig4');
%
% To display Fig. 5 use :
%
% exp_dietz2011('fig5');
%
% To display Fig. 6 use :
%
% exp_dietz2011('fig6');
%
% To display Fig. 6 from Spille et al. (2013) use :
%
% exp_dietz2011('fig6_spille2013');
%
% References:
% M. Dietz, S. D. Ewert, and V. Hohmann. Auditory model based direction
% estimation of concurrent speakers from binaural signals. Speech
% Communication, 53(5):592-605, 2011. [1]http ]
%
% C. Spille, B. Meyer, M. Dietz, and V. Hohmann. Binaural scene analysis
% with multi-dimensional statistical filters, chapter 6. Springer-Verlag
% GmbH, 2013.
%
% References
%
% 1. http://www.sciencedirect.com/science/article/pii/S016763931000097X
%
%
% Url: http://amtoolbox.sourceforge.net/amt-0.9.8/doc/experiments/exp_dietz2011.php
% Copyright (C) 2009-2015 Piotr Majdak and Peter L. Søndergaard.
% This file is part of AMToolbox version 0.9.8
%
% This program is free software: you can redistribute it and/or modify
% it under the terms of the GNU General Public License as published by
% the Free Software Foundation, either version 3 of the License, or
% (at your option) any later version.
%
% This program is distributed in the hope that it will be useful,
% but WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
% GNU General Public License for more details.
%
% You should have received a copy of the GNU General Public License
% along with this program. If not, see <http://www.gnu.org/licenses/>.
% AUTHOR: Mathias Dietz
definput.flags.type = {'missingflag','fig3','fig4','fig5','fig6','fig6_spille2013'};
definput.flags.plot = {'plot','noplot'};
[flags,keyvals] = ltfatarghelper({},definput,varargin);
if flags.do_missingflag
flagnames=[sprintf('%s, ',definput.flags.type{2:end-2}),...
sprintf('%s or %s',definput.flags.type{end-1},definput.flags.type{end})];
error('%s: You must specify one of the following flags: %s.',upper(mfilename),flagnames);
end;
% Load polynomial lookup data for converting ITD to azimuth
lookup = amt_load('dietz2011','itd2angle_lookuptable.mat');
if flags.do_fig3
signal=repmat(sig_competingtalkers('five_speakers'),2,[]);
fs = 16000;
signal=signal(1:5*fs,:);
s_pos = [-80 -30 0 30 80];
ic_threshold=0.98;
panellabel = 'ab';
% run IPD model on signal
[hairc_fine, fc, hairc_ild]=dietz2011(signal,fs,'fhigh',1400);
% convert interaural information into azimuth
itd_unwrapped = ...
dietz2011_unwrapitd(hairc_fine.itd_lp,hairc_ild,hairc_fine.f_inst,2.5);
angl=itd2angle(itd_unwrapped,lookup);
h_ic=zeros(91,12);
h_all=histc(angl,-90:2:90);
for n=1:12
h_ic(:,n)=histc(angl(hairc_fine.ic(:,n)>ic_threshold&[diff(hairc_fine.ic(:,n))>0; 0],n),-90:2:90);
end
example_output.angle_fine = angl;
example_output.IVS_fine = hairc_fine.ic;
example_output.histogram_angle_label = -90:2:90;
example_output.histograms_with_IVS = h_ic;
example_output.histograms_without_IVS = h_all;
if flags.do_plot
figure;
fontsize = 14;
set(gcf,'Position',[100 100 1170 700]);
for panel = 1:2
subplot(1,2,panel)
switch panel
case 1
bar(-90:2:90,sum(h_all,2),'r');
title('Mean histogram of all fine-structure channels','Fontsize',fontsize);
axis([-90 90 0 21900]);
set(gca,'YTick',[5000 10000 15000 20000],'YTickLabel',{'5k','10k','15k','20k'});
ymax = max(sum(h_all,2));
case 2
bar(-90:2:90,sum(h_ic,2));
title('Mean histogram with VS filter','Fontsize',fontsize);
axis([-90 90 0 5600]);
set(gca,'YTick',[1000:1000:5000],'YTickLabel',{'1k','2k','3k','4k','5k'});
ymax = max(sum(h_ic,2));
end
set(gca,'Fontsize',fontsize);
set(gca,'XTick',s_pos);
% xlim([-93 93])
% ylim([0 ymax*1.1])
xlabel('Azimuth [deg]','Fontsize',fontsize);
ylabel('Frequency of occurence','Fontsize',fontsize);
text (-80,ymax*.95,panellabel(panel),'Fontsize',fontsize+1,'FontWeight','bold');
end
end;
end;
if flags.do_fig4
% This reproduces Figure 4 from Dietz et al Speech Comm. 2011
signal=repmat(sig_competingtalkers('two_speakers'),2,[]);
fs = 16000;
s_pos = [-80 -30 0 30 80];
ic_threshold=0.98;
cn = [10 1]; % channel numbers for separate plots (1st entry also for time plot)
panellabel = 'abc';
% run IPD model on signal
[hairc_fine, fc, hairc_ild, hairc_mod216]=dietz2011(signal,fs,'mod_center_frequency_hz',216);
[~, ~, ~, hairc_mod135]=dietz2011(signal,fs,'mod_center_frequency_hz',135);
% convert interaural information into azimuth
itd_unwrapped = ...
dietz2011_unwrapitd(hairc_fine.itd_lp,hairc_ild(:,1:12),hairc_fine.f_inst,2.5);
angl=itd2angle(itd_unwrapped,lookup);
angl_fmod216=hairc_mod216.itd_lp*140000; %linear approximation. paper version is better than this
angl_fmod135=hairc_mod135.itd_lp*140000; %linear approximation. paper version is better than this
h_ic=zeros(61,12);
h_fmod216=zeros(61,11);
h_fmod135=zeros(61,11);
h_all=histc(angl,-60:2:60);
for n=1:12
h_ic(:,n)=histc(angl(hairc_fine.ic(:,n)>ic_threshold&[diff(hairc_fine.ic(:,n))>0; 0],n),-60:2:60);
end
for n=1:11
h_fmod216(:,n)=histc(angl_fmod216(hairc_mod216.ic(:,n)>ic_threshold&[diff(hairc_mod216.ic(:,n))>0; 0],n),-60:2:60);
h_fmod135(:,n)=histc(angl_fmod135(hairc_mod135.ic(:,n)>ic_threshold&[diff(hairc_mod135.ic(:,n))>0; 0],n),-60:2:60);
end
example_output.angle_fine = angl;
example_output.IVS_fine = hairc_fine.ic;
example_output.histogram_angle_label = -60:2:60;
example_output.histogram_panel1 = h_fmod216;
example_output.histogram_panel2 = h_fmod135;
example_output.histogram_panel3 = sum(h_ic,2);
if flags.do_plot
figure;
fontsize = 14;
set(gcf,'Position',[100 100 1170 400])
for panel = 1:3
subplot(1,3,panel)
switch panel
case 1
bar(-60:2:60,sum(h_fmod216,2))
title('histogram of mod ITD channels 13-23','Fontsize',fontsize)
axis([-50 50 0 7600]);
set(gca,'YTick',[2500 5000 7500],'YTickLabel',{'2.5k','5k','7.5k'});
ymax = max(sum(h_fmod216,2));
% ylim([0 ymax*1.15])
case 2
bar(-60:2:60,sum(h_fmod135,2))
axis([-50 50 0 7600]);
set(gca,'YTick',[2500 5000 7500],'YTickLabel',{'2.5k','5k','7.5k'});
title('histogram of mod ITD channels 13-23','Fontsize',fontsize)
ymax = max(sum(h_fmod135,2));
% ylim([0 ymax*1.15])
case 3
bar(-60:2:60,sum(h_ic,2))
title('histogram of fine ITD channels 1-12','Fontsize',fontsize)
axis([-50 50 0 76000]);
set(gca,'YTick',[25000 50000 75000],'YTickLabel',{'25k','50k','75k'});
ymax = max(sum(h_ic,2));
% ylim([0 ymax*1.15])
end
set(gca,'Fontsize',fontsize)
set(gca,'XTick',s_pos)
% xlim([-63 63])
xlabel('Azimuth [deg]','Fontsize',fontsize)
ylabel('Frequency of occurence','Fontsize',fontsize)
text (-40,ymax*1.2,panellabel(panel),'Fontsize',fontsize+1,'FontWeight','bold')
end
end;
end;
if flags.do_fig5
% mix signals
signal1=sig_competingtalkers('one_of_three');
signal2=sig_competingtalkers('two_of_three');
signal3=sig_competingtalkers('three_of_three');
noise =sig_competingtalkers('bnoise');
noise = noise(1:40000,:);
fs = 16000;
% derive histograms
ic_threshold=0.98;
k = zeros(14,38);
for n = 1:7
switch n
case 1
signal = signal1+noise;
case 2
signal = signal2+noise;
case 3
signal = signal3+noise;
case 4
signal = signal1+2*noise;
case 5
signal = signal2+2*noise;
case 6
signal = signal3+2*noise;
case 7
signal = noise;
end
% run IPD model on signal
[hairc_fine, fc, hairc_ild]=dietz2011(signal,fs,'fhigh',1400);
% convert interaural information into azimuth
itd_unwrapped = ...
dietz2011_unwrapitd(hairc_fine.itd_lp,hairc_ild,hairc_fine.f_inst,2.5);
angl=itd2angle(itd_unwrapped,lookup);
h_ic=zeros(38,12);
k(2*n,:)=sum(histc(angl,-92.5:5:92.5),2);
for erb=1:12
h_ic(:,erb)=histc(angl(hairc_fine.ic(:,erb)>ic_threshold&[diff(hairc_fine.ic(:,erb))>0; 0],erb),-92.5:5:92.5);
end
k(2*n-1,:)=sum(h_ic,2);
example_output.angle_fine(:,:,n)=angl;
example_output.IVS_fine = hairc_fine.ic;
example_output.histogram_angle_label = -92.5:5:92.5;
example_output.histograms_with_IVS(:,n)=sum(h_ic,2);
example_output.histograms_without_IVS(:,n)=sum(histc(angl,-92.5:5:92.5),2);
end
if flags.do_plot
% plot
figure;
set(gcf,'Position',[100 100 990 500])
y=[0.14 0.57];
x=[0.1 0.22 0.34 0.49 0.61 0.73 0.88];
cols = 2;
condi={'1S 0dB','','2S 0dB','',...
'3S 0dB','','1S -6dB','',...
'2S -6dB','','3S -6dB','','noise',''};
for n = 1:14
axes('position',[x(ceil(n/cols)) y(mod(n-1,cols)+1) 0.11 0.42],...
'box','on','fontSize',12)
if mod(n,2)==0
set(gca,'YTick',[10 20 30],'YTickLabel',{'','',''},'ylim',[0 32])
set(gca,'Visible','on','XTick',[-30 0 30],...
'xlim',[-50 50],'XTickLabel',{'','','',''})
else
set(gca,'YTick',[1 2 3],'YTickLabel',{'','',''},'ylim',[0 3.33])
set(gca,'Visible','on','XTick',[-30 0 30],...
'xlim',[-50 50],'XTickLabel',{'-30','0','+30'})
xlabel(condi(n))
end
if n==2
set(gca,'YTick',[10 20 30],'YTickLabel',{'10k','20k','30k'})
ylabel('no IVS mask')
elseif n==1
set(gca,'YTick',[1 2 3],'YTickLabel',{'1k','2k','3k'})
ylabel('with IVS mask')
end
hold on
bar(-92.5:5:92.5,k(n,:)/1000)
colormap gray
end
hold off
end;
end;
if flags.do_fig6
signal=sig_competingtalkers('one_speaker_reverb');
fs = 16000;
s_pos = [-45 0 45];
ic_threshold=0.98;
cn = [10 1]; % channel numbers for separate plots (1st entry also for time plot)
panellabel = 'abcd';
% run IPD model on signal
[hairc_fine, fc, hairc_ild, hairc_mod]=dietz2011(signal,fs);
% convert interaural information into azimuth
itd_unwrapped = ...
dietz2011_unwrapitd(hairc_fine.itd_lp,hairc_ild(:,1:12),hairc_fine.f_inst,2.5);
angl=itd2angle(itd_unwrapped,lookup);
angl_fmod=hairc_mod.itd_lp*140000; %linear approximation. paper version is better than this
h_ic=zeros(71,12);
h_fmod=zeros(71,11);
h_all=histc(angl,-70:2:70);
for n=1:12
h_ic(:,n)=histc(angl(hairc_fine.ic(:,n)>ic_threshold&[diff(hairc_fine.ic(:,n))>0; 0],n),-70:2:70);
end
for n=1:11
h_fmod(:,n)=histc(angl(hairc_mod.ic(:,n)>ic_threshold&[diff(hairc_mod.ic(:,n))>0; 0],n),-70:2:70);
end
example_output.angle_fine = angl;
example_output.IVS_fine = hairc_fine.ic;
example_output.histogram_angle_label = -70:2:70;
example_output.histograms_with_IVS = h_ic;
example_output.histograms_without_IVS = h_all;
example_output.histogram_panel1 = sum(h_ic,2);
example_output.histogram_panel2 = sum(h_ic(:,6:12),2);
example_output.histogram_panel3 = sum(h_ic(:,1:3),2);
example_output.histogram_panel4 = sum(h_fmod,2);
if flags.do_plot
figure;
fontsize = 14;
set(gcf,'Position',[60 100 1370 350])
for panel = 1:4
subplot(1,4,panel)
switch panel
case 1
bar(-70:2:70,sum(h_ic,2))
title('fine ITD channels 200-1400 Hz','Fontsize',fontsize)
ymax = max(sum(h_ic,2));
case 2
bar(-70:2:70,sum(h_ic(:,6:12),2))
title('fine ITD channels 500-1400 Hz','Fontsize',fontsize)
ymax = max(sum(h_ic,2));
case 3
bar(-70:2:70,sum(h_ic(:,1:3),2))
title('fine ITD channels 200-400 Hz','Fontsize',fontsize)
ymax = max(sum(h_ic,2));
case 4
bar(-70:2:70,sum(h_fmod,2))
title('mod ITD channels 1400-5000 Hz','Fontsize',fontsize)
ymax = max(sum(h_ic,2));
end
set(gca,'Fontsize',fontsize)
set(gca,'XTick',s_pos)
% xlim([-73 73])
% ylim([0 ymax*1.1])
axis([-73 73 0 5500]);
set(gca,'YTick',[1000:1000:5000],'YTickLabel',{'1k','2k','3k','4k','5k'});
xlabel('Azimuth [deg]','Fontsize',fontsize)
ylabel('Frequency of occurence','Fontsize',fontsize)
text (-50,ymax*.97,panellabel(panel),'Fontsize',fontsize+1,'FontWeight','bold')
end
end;
end;
if flags.do_fig6_spille2013
signal=amt_load('dietz2011','s123456.wav');
fs=44100;
s_pos =[-75 -40 -10 10 40 75];
ic_threshold=0.98;
cn = [10 1]; % channel numbers for separate plots (1st entry also for time plot)
panellabel = 'acbd';
% run IPD model on signal
[hairc_fine,fc,hairc_ild]=dietz2011(signal,fs,'fhigh',1400);
% convert interaural information into azimuth
itd_unwrapped = ...
dietz2011_unwrapitd(hairc_fine.itd_lp,hairc_ild,hairc_fine.f_inst_lp,2.5);
lookup = amt_load('dietz2011','itd2angle_lookuptable.mat');
angl=itd2angle(itd_unwrapped,lookup);
h_ic=zeros(91,12);
h_all=histc(angl,-90:2:90);
for n=1:12
h_ic(:,n)=histc(angl(hairc_fine.ic(:,n)>ic_threshold&[diff(hairc_fine.ic(:,n))>0; 0],n),-90:2:90);
end
% plot
figure;
fontsize = 14;
set(gcf,'Position',[100 100 1170 700])
for panel = 1:4
subplot(2,2,panel)
switch panel
case 1
bar(-90:2:90,h_all(:,cn(1)))
hold on
bar(-90:2:90,h_ic(:,cn(1)),'r')
title(['Azimuth histogram of ch. ' num2str(cn(1)) ' at cf = ' ...
num2str(round(fc(cn(1)))) ' Hz'],'Fontsize',fontsize)
ymax = max(h_all(:,cn(1)));
case 2
bar(-90:2:90,h_all(:,cn(2)))
hold on
bar(-90:2:90,h_ic(:,cn(2)),'r')
title(['Azimuth histogram of ch. ' num2str(cn(2)) ' at cf = ' ...
num2str(round(fc(cn(2)))) ' Hz'],'Fontsize',fontsize)
ymax = max(h_all(:,cn(2)));
case 3
t=(1:length(signal))*1/fs;
plot(t,angl(:,cn(1)),'b.');
hold on;
plot(t(hairc_fine.ic(:,cn(1))>ic_threshold),...
angl(hairc_fine.ic(:,cn(1))>ic_threshold,cn(1)),'r.');
title(['Azimuth over time in ch. ' num2str(cn(1)) ' at cf = ' ...
num2str(round(fc(cn(1)))) ' Hz'],'Fontsize',fontsize)
case 4
bar(-90:2:90,mean(h_all,2))
hold on
bar(-90:2:90,mean(h_ic,2),'r')
title('Mean histogram of all fine-structure channels','Fontsize',fontsize)
ymax = max(mean(h_all,2));
end
set(gca,'Fontsize',fontsize)
if panel ~= 3
set(gca,'XTick',s_pos)
xlim([-93 93])
ylim([0 ymax*1.1])
xlabel('Azimuth [deg]','Fontsize',fontsize)
ylabel('Frequency of occurence','Fontsize',fontsize)
text (-80,ymax*.95,panellabel(panel),'Fontsize',fontsize+1,'FontWeight','bold')
else
set(gca,'YTick',s_pos)
xlim([0 5.5])
ylim([-95 95])
ylabel('Azimuth [deg]','Fontsize',fontsize)
xlabel('Time [s]','Fontsize',fontsize)
rectangle('Position',[0.32,60,0.3,21],'FaceColor','white')
text (0.38,69,panellabel(panel),'Fontsize',fontsize+1,'FontWeight','bold')
end
end
example_output.angle_fine = angl;
example_output.IVS_fine = hairc_fine.ic;
example_output.histogram_angle_label = -90:2:90;
example_output.histograms_with_IVS = h_ic;
example_output.histograms_without_IVS = h_all;
end