THE AUDITORY MODELING TOOLBOX
Go to function
EXP_BAUMGARTNER2015 - Results from Baumgartner and Majdak (2015)
Program code:
function varargout = exp_baumgartner2015(varargin)
%EXP_BAUMGARTNER2015 Results from Baumgartner and Majdak (2015)
% Usage: data = exp_baumgartner2015(flag)
%
% EXP_BAUMGARTNER2015(flag) reproduces figures of the studies from
% Baumgartner et al. (2015).
%
%
% The following flags can be specified
%
%
% 'fig2' Reproduce Fig.2 of Baumgartner and Majdak (2015):
% Example showing the spectral discrepancies obtained by VBAP.
% The targeted spectrum is the HRTF for 20 deg polar angle.
% The spectrum obtained by VBAP is the superposition of two
% HRTFs from directions 40 deg polar angle apart of each
% other with the tar- geted source direction centered in between.
%
% 'fig4' Reproduce Fig.4 of Baumgartner and Majdak (2015):
% Response predictions to sounds created by VBAP with two
% loudspeakers in the median plane positioned at polar
% angles of -15 and 30 deg, respectively. Predictions for
% two exemplary listeners and pooled across all listeners.
% Each column of a panel shows the predicted PMV of
% polar-angle responses to a certain sound. Note the
% inter-individual differences and the generally small
% probabilities at response angles not occupied by the loudspeakers.
%
% 'fig5' Reproduce Fig.5 of Baumgartner and Majdak (2015):
% Listener-specific increases in polar error as a function of
% the panning angle. Increase in polar error defined as
% the difference between the polar error obtained by the
% VBAP source and the polar error obtained by the real
% source at the corresponding panning angle. Same loudspeaker
% arrangement as for Fig. 4. Note the large inter-individual
% differences and the increase in polar error being largest
% at panning angles centered between the loudspeakers, i.e.,
% at panning ratios around R = 0 dB.
%
% 'fig6' Reproduce Fig.6 of Baumgartner and Majdak (2015):
% Panning angles for the loudspeaker arrangement of Fig. 4
% judged best for reference sources at polar angles of
% 0 or 15 deg in the median plane. Comparison between
% experimental results from [2] and simulated results
% based on various response strategies: PM, CM, and both
% mixed. Dotted horizontal line: polar angle of the reference
% source. Hor- izontal line within box: median;
% box: inter-quartile range (IQR);
% whisker: within quartile +-1.5 IQR;
% star: outlier.
% Note that the simulations predicted a bias similar to
% the results from Pulkki (2001) for the reference source at 0 deg.
%
% 'tab1' Reproduce Tab.1 of Baumgartner and Majdak (2015):
% Means and standard deviations of responded panning angles for the
% two reference sources (Ref.) together with corresponding GOFs
% evaluated for the actual results from Pulkki (2001) and
% predicted results based on various response strategies.
% Note the relatively large GOFs for the simulations based on
% mixed response strategies indicating a good correspondence
% between actual and predicted results.
%
% 'fig7' Reproduce Fig.7 of Baumgartner and Majdak (2015):
% Increase in polar error (defined as in Fig. 5) as a function
% of loudspeaker span in the median plane with panning ratio
% R = 0 dB. Black line with gray area indicates mean
% +-1 standard deviation across listeners. Note that the
% increase in polar error monotonically increases with
% loudspeaker span.
%
% 'fig8' Reproduce Fig.8 of Baumgartner and Majdak (2015):
% Effect of loudspeaker span in the median plane on coefficient
% of determination, r^2, for virtual source directions
% created by VBAP. Separate analysis for frontal, rear,
% and overall (frontal and rear) targets. Data pooled
% across listeners. Note the correspondence with the
% results obtained by Bremen et al. (2010).
%
% 'tab3' Reproduce Tab.3 of Baumgartner and Majdak (2015):
% Predicted across-listener average of increase in polar
% errors as referred to a reference system containing
% loudspeakers at all considered directions. Distinction
% between mean and maximum degradation across directions.
% N: Number of loudspeakers. Ele.: Elevation of second layer.
% Notice that this elevation has a larger effect on mean
% and maximum degradation than N.
%
% 'fig9' Reproduce Fig.9 of Baumgartner and Majdak (2015):
% Predicted polar error as a function of the lateral and
% polar angle of a virtual source created by VBAP in
% various multichannel systems. Open circles indicate
% loudspeaker directions. Reference shows polar error
% predicted for a real source placed at the virtual
% source directions investigated for systems A, ..., F.
%
%
% Further, cache flags (see amt_cache) and plot flags can be specified:
%
% 'plot' Plot the output of the experiment. This is the default.
%
% 'no_plot' Don't plot, only return data.
%
% Requirements:
% -------------
%
% 1) SOFA API v0.4.3 or higher from http://sourceforge.net/projects/sofacoustics for Matlab (in e.g. thirdparty/SOFA)
%
% 2) Data in hrtf/baumgartner2014
%
% 3) Statistics Toolbox for Matlab (for some of the figures)
%
% Examples:
% ---------
%
%
% To display Fig.2 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('fig2');
%
% To display Fig.4 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('fig4');
%
% To display Fig.5 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('fig5');
%
% To display Fig.6 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('fig6');
%
% To display Fig.7 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('fig7');
%
% To display Fig.8 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('fig8');
%
% To display Fig.9 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('fig9');
%
% To display Tab.1 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('tab1');
%
% To display Tab.3 of Baumgartner and Majdak (2015) use :
%
% exp_baumgartner2015('tab3');
%
% See also: baumgartner2014 data_baumgartner2014
%
% References:
% R. Baumgartner, P. Majdak, and B. Laback. Modeling sound-source
% localization in sagittal planes for human listeners. The Journal of the
% Acoustical Society of America, 136(2):791--802, 2014.
%
% R. Baumgartner and P. Majdak. Modeling Localization of Amplitude-Panned
% Virtual Sources in Sagittal Planes. J. Audio Eng. Soc.,
% 63(7/8):562--569, Aug. 2015. [1]http ]
%
% P. Bremen, M. M. van Wanrooij, and A. J. van Opstal. Pinna cues
% determine orientation response modes to synchronous sounds in
% elevation. J. Neurosci., 30 (1):194--204, 2010.
%
% V. Pulkki. Localization of Amplitude-Panned Virtual Sources II: Two-
% and Three-Dimensional Panning. J. Audio Eng. Soc., 49(4):753--767,
% 2001.
%
% References
%
% 1. http://www.aes.org/e-lib/browse.cfm?elib=17842
%
%
% Url: http://amtoolbox.org/amt-1.5.0/doc/experiments/exp_baumgartner2015.php
% #Author: Robert Baumgartner (2015)
% This file is licensed unter the GNU General Public License (GPL) either
% version 3 of the license, or any later version as published by the Free Software
% Foundation. Details of the GPLv3 can be found in the AMT directory "licences" and
% at <https://www.gnu.org/licenses/gpl-3.0.html>.
% You can redistribute this file and/or modify it under the terms of the GPLv3.
% This file is distributed without any warranty; without even the implied warranty
% of merchantability or fitness for a particular purpose.
%% ------ Check input options --------------------------------------------
definput.import={'amt_cache'};
definput.keyvals.FontSize = 12;
definput.keyvals.MarkerSize = 6;
% definput.flags.type = {'missingflag',...% 'fig5_baumgartner2015aro',...
% 'fig2_baumgartner2015jaes','fig4_baumgartner2015jaes',...
% 'fig5_baumgartner2015jaes','fig6_baumgartner2015jaes',...
% 'fig7_baumgartner2015jaes','fig8_baumgartner2015jaes',...
% 'fig9_baumgartner2015jaes','tab1_baumgartner2015jaes',...
% 'tab3_baumgartner2015jaes',...
% };
definput.flags.type = {'missingflag',...
'fig2','fig4',...
'fig5','fig6',...
'fig7','fig8',...
'fig9','tab1',...
'tab3',...
};
definput.flags.plot = {'plot','no_plot'};
[flags,kv] = ltfatarghelper({'FontSize','MarkerSize'},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;
%% General Plot Settings
TickLength = [0.02,0.04];
%% ------------------------------------------------------------------------
% ---- baumgartner2015jaes -----------------------------------------------
% ------------------------------------------------------------------------
%% ------ FIG 2 of baumgartner2015jaes ------------------------------------
if flags.do_fig2
pol1 = 0;
pol2 = 40;
s = data_baumgartner2014('pool');
fs = s(1).Obj.Data.SamplingRate;
[dtfs,pol] = extractsp(0,s(1).Obj);
polphant = pol1 + (pol2-pol1)/2;
dtf1 = 10*dtfs(:,pol==pol1,1);
dtf2 = 10*dtfs(:,pol==pol2,1);
dtfreal = 10*dtfs(:,pol==polphant,1);
dtfphant = (dtf1+dtf2)/2;
%%
if flags.do_plot
figure
set(gca,'LineWidth',1)
plotfft(fft(dtfphant),fs,'posfreq')
hold on
plotfft(fft(dtfreal),fs,'posfreq')
plotfft(fft(dtf1),fs,'posfreq')
plotfft(fft(dtf2),fs,'posfreq');
% Set line styles
Color = {[0.4660 0.6740 0.1880];...
[0.3010 0.7450 0.9330];...
[0.8500 0.3250 0.0980];...
[ 0 0.4470 0.7410]};
LineWidth = [.5,.5,1,1];
LineStyle = {':',':','--','-'};
ch = get(gca,'Children');
for ii = 1:length(ch)
set(ch(ii),'Color',Color{ii},'LineStyle',LineStyle{ii},'LineWidth',LineWidth(ii));
end
leg = legend([num2str(polphant) '\circ VBAP'],...
[num2str(polphant) '\circ source'],...
[' ' num2str(pol1) '\circ source'],...
[num2str(pol2) '\circ source']);
set(leg,'Location','southeast','FontSize',kv.FontSize)
set(gca,'XLim',[500,17500],'YLim',[-39.9,9.9],'FontSize',kv.FontSize)
xticks = get(gca,'XTick');
set(gca,'XTickLabel',xticks/1000)
xlabel({'Frequency (kHz)';' '},'FontSize',kv.FontSize)
end
end
%% ------ FIG 4 of baumgartner2015jaes ------------------------------------
if flags.do_fig4
[peI,s,pol0,DL,respang] = amt_cache('get','panningangle',flags.cachemode);
if isempty(peI)
MRS = 0;
pol1 = -15; % polar angle of lower Lsp.
pol2 = 30; % polar angle of higher Lsp.
lat = 0; % must be 0 otherwise VBAP wrong!
s = data_baumgartner2014('pool');
dPol = pol2-pol1;
s(1).spdtfs = [];
[s(1).spdtfs,polang] = extractsp(lat,s(1).Obj);
idtest = find(polang >= pol1 & polang <= pol2);
qeI = zeros(length(s),length(idtest));
peI = qeI;
for ii = 1:length(idtest)
id1 = idtest(1); % ID lower pos.
id2 = idtest(end); % ID higher pos.
id0 = idtest(ii); % ID pos. of phantom source
pol0(ii) = polang(id0);
% VBAP
[p(1,1),p(1,2),p(1,3)] = sph2cart(lat,deg2rad(pol0(ii)),1);
[L(1,1),L(1,2),L(1,3)] = sph2cart(lat,deg2rad(pol1),1);
[L(2,1),L(2,2),L(2,3)] = sph2cart(lat,deg2rad(pol2),1);
g = p/L;
g = g/norm(g);
DL(ii) = db(g(1)) - db(g(2));
for ll = 1:length(s)
s(ll).spdtfs = extractsp(lat,s(ll).Obj);
% superposition
s(ll).dtfs2{ii} = g(1)*s(ll).spdtfs(:,id1,:) + g(2)*s(ll).spdtfs(:,id2,:);
[s(ll).p1(:,ii),respang] = baumgartner2014(...
s(ll).spdtfs(:,id0,:),s(ll).spdtfs,s(ll).fs,'S',s(ll).S,...
'mrsmsp',MRS,'polsamp',polang);
s(ll).p2(:,ii) = baumgartner2014(...
s(ll).dtfs2{ii},s(ll).spdtfs,s(ll).fs,'S',s(ll).S,...
'mrsmsp',MRS,'polsamp',polang);
[s(ll).qe1(ii),s(ll).pe1(ii)] = baumgartner2014_pmv2ppp(...
s(ll).p1(:,ii),polang(id0),respang);
[s(ll).qe2(ii),s(ll).pe2(ii)] = baumgartner2014_pmv2ppp(...
s(ll).p2(:,ii),polang(id0),respang);
% Increse of error
qeI(ll,ii) = s(ll).qe2(ii) - s(ll).qe1(ii);
peI(ll,ii) = s(ll).pe2(ii) - s(ll).pe1(ii);
end
amt_disp([num2str(ii) ' of ' num2str(length(idtest)) ' completed']);
end
amt_cache('set','panningangle',peI,s,pol0,DL,respang);
end
id1 = 'NH71'; % positive example
id2 = 'NH62'; % negative example
if flags.do_plot
cmp= [0.2081 0.1663 0.5292;
0.2052 0.2467 0.6931;
0.0843 0.3472 0.8573;
0.0157 0.4257 0.8789;
0.0658 0.4776 0.8532;
0.0777 0.5300 0.8279;
0.0356 0.5946 0.8203;
0.0230 0.6443 0.7883;
0.0485 0.6793 0.7341;
0.1401 0.7085 0.6680;
0.2653 0.7327 0.5916;
0.4176 0.7471 0.5142;
0.5624 0.7487 0.4529;
0.6872 0.7433 0.4029;
0.7996 0.7344 0.3576;
0.9057 0.7261 0.3105;
0.9944 0.7464 0.2390;
0.9847 0.8141 0.1734;
0.9596 0.8869 0.1190;
0.9763 0.9831 0.0538]; % parula colormap (defined for compatibility with older Matlab versions)
figure
p_pool = nan(size(s(1).p2,1),size(s(1).p2,2),length(s));
for ll = 1:length(s)
p_pool(:,:,ll) = s(ll).p2;
if strcmp(s(ll).id,id1)
subplot(1,4,1)
plot_baumgartner2014(s(ll).p2,pol0,respang,'cmax',0.08)
colormap(cmp)
title(s(ll).id,'FontSize',kv.FontSize)
xlabel('');
ylabel('Response angle (deg)','FontSize',kv.FontSize);
colorbar off
elseif strcmp(s(ll).id,id2)
subplot(1,4,2)
plot_baumgartner2014(s(ll).p2,pol0,respang,'cmax',0.08)
colormap(cmp)
xlabel('Panning angle (deg)','FontSize',kv.FontSize);
ylabel(''); set(gca,'YTickLabel',[]);
title(s(ll).id,'FontSize',kv.FontSize)
colorbar off
end
end
p_pool = mean(p_pool,3);
subplot(1,4,3)
plot_baumgartner2014(p_pool,pol0,respang,'cmax',0.08)
colormap(cmp)
title('Pool','FontSize',kv.FontSize)
xlabel('');
ylabel(''); set(gca,'YTickLabel',[]);
colorbar off
subplot(1,4,4)
ymax = 8;
y = -0.15:0.1:ymax+0.15;
Ny = length(y);
pcolor(1:2,y,repmat(y(:),1,2))
shading flat
axis tight
colormap(cmp)
title({' ';' '})
set(gca,'XTick',[],'YTick',0:1:ymax,...
'YDir','normal','YAxisLocation','right','FontSize',kv.FontSize)
ylabel({'Probability density (% per 5\circ)'},'FontSize',kv.FontSize)
set(gca,'Position',get(gca,'Position').*[1.05,1,0.3,1])
end
end
%% ------ FIG 5 of baumgartner2015jaes ------------------------------------
if flags.do_fig5
[peI,s,pol0,DL,respang] = amt_cache('get','panningangle',flags.cachemode);
if isempty(peI)
exp_baumgartner2015('fig4','no_plot',flags.cachemode);
[peI,s,pol0,DL,respang] = amt_cache('get','panningangle',flags.cachemode);
end
figure
plot(peI')
ylabel('Increase in polar error (deg)')
% Panning angle axis
set(gca,'XTick',1:10,'XTickLabel',round(pol0),...
'YMinorTick','on','XLim',[1,10],'YLim',[-9,59])
xlabel('Panning angle (deg)')
end
%% ------ FIG 6 of baumgartner2015jaes ------------------------------------
if flags.do_fig6
results = amt_cache('get','replicatePulkki2001',flags.cachemode);
if isempty(results)
amt_disp('Results may slightly vary from simulation to simulation because noise stimulus is not fixed.');
MRS = 0;
pol1 = -15; % polar angle of lower Lsp.
pol2 = 30; % polar angle of higher Lsp.
polphant = [0,15]; % polar angles of phantom sources
Pmax = nan(length(polphant),23); % max Probabilities
panang_Pmax = nan(length(polphant),23); % panning angle selected by max P
panang_Cen = nan(length(polphant),23); % panning angle selected by centroid
for pp = 1:length(polphant)
lat = 0; % must be 0 otherwise VBAP wrong!
s = data_baumgartner2014('pool');
s(1).spdtfs = [];
[s(1).spdtfs,polang] = extractsp(lat,s(1).Obj);
% restrict response range
idrang = find(polang >= pol1 & polang <= pol2); % to range between loudspeakers
% idrang = find(polang <= 90); % to the front
idtest = find(polang >= pol1 & polang <= pol2);
id1 = find(polang >= pol1,1); % ID lower pos.
id2 = find(polang <= pol2,1,'last'); % ID higher pos.
tang = polang(id1:id2);
for ll = 1:length(s)
s(ll).spdtfs = extractsp(lat,s(ll).Obj);
for ii = 1:length(idtest)
id0 = idtest(ii); % ID pos. of phantom source
% VBAP
[p(1,1),p(1,2),p(1,3)] = sph2cart(lat,deg2rad(polang(id0)),1);
[L(1,1),L(1,2),L(1,3)] = sph2cart(lat,deg2rad(pol1),1);
[L(2,1),L(2,2),L(2,3)] = sph2cart(lat,deg2rad(pol2),1);
g = p/L;
g = g/norm(g);
DL(ii) = db(g(1)) - db(g(2));
% superposition
s(ll).dtfs2{ii} = g(1)*s(ll).spdtfs(:,id1,:) + g(2)*s(ll).spdtfs(:,id2,:);
[s(ll).p(:,ii),rang] = baumgartner2014(...
s(ll).dtfs2{ii},s(ll).spdtfs(:,idrang,:),s(ll).fs,'S',s(ll).S,...
'mrsmsp',MRS,'polsamp',polang(idrang),'rangsamp',5,...
'stim',noise(10000,1,'pink')); % phantom source
end
id_rang = rang == polphant(pp);
% interpolation between target angles
tang_int = tang(1):1:tang(end);
p_int = interp2(tang(:)',rang(:),s(ll).p,tang_int(:)',rang(:),'spline');
p_int = p_int./repmat(sum(p_int,1),size(p_int,1),1); % normalize to PMVs
% Variant 1: max P at source direction
[Pmax(pp,ll),id_best_pan] = max(p_int(id_rang,:));
panang_Pmax(pp,ll) = tang_int(id_best_pan);
% Variant 2: centroid closest to source direction
M = rang*p_int;
[tmp,id_best_pan] = min(abs(M-polphant(pp)));
panang_Cen(pp,ll) = tang_int(id_best_pan);
end
fprintf([num2str(pp) ' of ' num2str(length(polphant)) ' completed \n']);
end
results = struct('panang_Pmax',panang_Pmax,'Pmax',Pmax,...
'panang_Cen',panang_Cen,'polphant',polphant,'DL',DL,'rang',rang);
amt_cache('set','replicatePulkki2001',results);
end
[panang_varStrat,nCM_varStrat,p_varStrat,muhat,sigmahat] = amt_cache('get','replicatePulkki2001_varStrat',flags.cachemode);
if isempty(panang_varStrat)
pulkki01 = data_pulkki2001;
[muhat(1),sigmahat(1)] = normfit(pulkki01(1,:));
[muhat(2),sigmahat(2)] = normfit(pulkki01(2,:));
Nsub = size(results.panang_Cen,2);
panang_all = [];
p = [];
nCM = [];
tmp = [];
ii = 1;
for inCM = 1:Nsub+1
c = nchoosek(1:Nsub,inCM-1); % listeners with CM strategy
lenC = size(c,1);
nCM = [nCM;repmat(inCM,lenC,1)];
panang_all = cat(3,panang_all , nan(2,Nsub,lenC));
p = [p ; nan(lenC,2)];
for ic = 1:lenC
idCM = false(1,Nsub);
idCM(c(ic,:)) = true;
panang_all(:,:,ii) = [results.panang_Cen(:,idCM) results.panang_Pmax(:,not(idCM))];
[tmp.h1,p(ii,1)] = kstest((panang_all(1,:,ii)-muhat(1))/sigmahat(1)); % center data acc. to target distribution and then test similarity to standard normal distribution
[tmp.h2,p(ii,2)] = kstest((panang_all(2,:,ii)-muhat(2))/sigmahat(2));
ii = ii+1;
end
disp([num2str(inCM) ' of ' num2str(Nsub+1) ' done'])
end
[tmp.min,idmax] = max(sum(p,2)); % best fit
panang_varStrat = panang_all(:,:,idmax);
nCM_varStrat = nCM(idmax);
p_varStrat = p(idmax,:);
amt_cache('set','replicatePulkki2001_varStrat',panang_varStrat,nCM_varStrat,p_varStrat,muhat,sigmahat)
end
if flags.do_plot
pulkki01 = data_pulkki2001;
Nsub = size(results.panang_Cen,2);
figure
for ii = 1:2
subplot(1,2,ii)
X = nan(size(pulkki01,2)*size(pulkki01,3),4);
X(:,1) = pulkki01(ii,:);
X(1:Nsub,2) = results.panang_Pmax(ii,:)';
X(1:Nsub,3) = results.panang_Cen(ii,:)';
X(1:Nsub,4) = panang_varStrat(ii,:)';
plot([0,5],(ii-1)*15*[1,1],'k:')
hold on
boxplot(X,'symbol','k*','outliersize',3)
set(gca,'YLim',[-17,32], 'XTickLabel',{'[2]','PM','CM','Mixed'});
if ~verLessThan('matlab','8.4'), set(gca,'XTickLabelRotation',45); end
if ii==1;
ylabel('Panning angle (deg)')
text(0.7,27.5,'0\circ')
else
set(gca,'YTickLabel',[]);
text(0.7,27.5,'15\circ')
end
end
end
end
%% ------ FIG 7 of baumgartner2015jaes ------------------------------------
if flags.do_fig7
[peI,dPol] = amt_cache('get','loudspeakerspan',flags.cachemode);
if isempty(peI)
MRS = 0;
flags.do_fig20 = false;
flags.do_fig19 = false;
s = data_baumgartner2014('pool');
if flags.do_fig19
dPol = [0 30,60];
s = s(2); % NH12
else
dPol = 10:10:90;
end
lat = 0;
s(1).spdtfs = [];
[s(1).spdtfs,polang] = extractsp(lat,s(1).Obj);
peI = zeros(length(s),length(dPol));
peA = zeros(length(s),length(dPol)+1);
ii = 0;
while ii < length(dPol)
ii = ii + 1;
% find comparable angles
id0 = [];
id1 = [];
id2 = [];
for jj = 1: length(polang)
t0 = find( round(polang) == round(polang(jj)+dPol(ii)/2) );
t2 = find( round(polang) == round(polang(jj)+dPol(ii)) );
if ~isempty(t0) && ~isempty(t2)
id0 = [id0 t0];
id1 = [id1 jj];
id2 = [id2 t2];
end
end
pol2{ii} = (polang(id1)+polang(id2)) /2;
amt_disp([' Span: ' num2str(dPol(ii)) 'deg']);
for ll = 1:length(s)
s(ll).spdtfs = extractsp(lat,s(ll).Obj);
% superposition
s(ll).dtfs2{ii} = s(ll).spdtfs(:,id1,:) + s(ll).spdtfs(:,id2,:);
[s(ll).p1{ii},respang] = baumgartner2014(...
s(ll).spdtfs(:,id0,:),s(ll).spdtfs,s(ll).fs,'S',s(ll).S,...
'mrsmsp',MRS,'polsamp',polang);
s(ll).p2{ii} = baumgartner2014(...
s(ll).dtfs2{ii},s(ll).spdtfs,s(ll).fs,'S',s(ll).S,...
'mrsmsp',MRS,'polsamp',polang);
% RMS Error
[s(ll).qe1{ii},s(ll).pe1{ii}] = baumgartner2014_pmv2ppp(...
s(ll).p1{ii},polang(id0),respang);
[s(ll).qe2{ii},s(ll).pe2{ii}] = baumgartner2014_pmv2ppp(...
s(ll).p2{ii},pol2{ii},respang);
% Increse of error
peI(ll,ii) = s(ll).pe2{ii} - s(ll).pe1{ii};
end
end
amt_cache('set','loudspeakerspan',peI,dPol)
end
peIm = mean(peI,1);
peIstd = std(peI,1,1);
if flags.do_plot
figure
p = patch([dPol,fliplr(dPol)],[peIm+peIstd,fliplr(peIm-peIstd)],.7*[1,1,1]);
set(p,'EdgeColor',.7*[1,1,1]);
hold on
h = plot(dPol,peIm,'k');
set(h,'LineWidth',1)
set(gca,'XTick',dPol,'XTickLabel',dPol,...
'YMinorTick','on','Box','on','Layer','top')
axis([9.8,90.2,-2,27])
ylabel({'Increase in polar error (deg)'})
xlabel({'Loudspeaker span (deg)';''})
end
end
%% ------ FIG 8 of baumgartner2015jaes ------------------------------------
if flags.do_fig8
[r2,dPol] = amt_cache('get','loudspeakerspan_r2',flags.cachemode);
if isempty(r2)
MRS = 0;
s = data_baumgartner2014('pool');
dPol = 0:10:105;
lat = 0;
runs = 100;
DL = -13:5:7; % panning ratios in dB
s(1).spdtfs = [];
[s(1).spdtfs,polang] = extractsp(lat,s(1).Obj);
r2 = zeros(length(s),length(dPol));
r2 = [];
ii = 0;
while ii < length(dPol) % various spans
ii = ii + 1;
disp([' Span: ' num2str(dPol(ii)) 'deg']);
r2total = nan(length(s),length(DL));
r2front = nan(length(s),length(DL));
r2rear = nan(length(s),length(DL));
for idl = 1:length(DL)
% find comparable angles
id1 = []; % id of speaker with smaller polar angle
id2 = []; % id of speaker with larger polar angle
for jj = 1: length(polang)
% t0 = find( round(polang) == round(polang(jj)+ipf*dPol(ii)/polFrac) );
t2 = find( round(polang) == round(polang(jj)+dPol(ii)) );
if ~isempty(t2)
% id0 = [id0 t0];
id1 = [id1 jj];
id2 = [id2 t2];
end
end
g = inv([1,1;1,-10^(DL(idl)/20)]) * [1;0]; % derived from db(g1/g2)=DL and g1+g2=1 (chosen arbitrarily since energy preservation is not relevant here)
pol0 = nan(length(id1),1); % panning angles
for jj = 1:length(id1)
[L(1,1),L(1,2),L(1,3)] = sph2cart(0,deg2rad(polang(id1(jj))),1);
[L(2,1),L(2,2),L(2,3)] = sph2cart(0,deg2rad(polang(id2(jj))),1);
t = g'*L;
[azi,ele,tmp.r] = cart2sph(t(1),t(2),t(3));
[tmp.lat,pol0(jj)] = sph2hor(rad2deg(azi),rad2deg(ele));
end
for ll = 1:length(s) % various listeners
s(ll).spdtfs = extractsp(lat,s(ll).Obj);
% superposition
s(ll).dtfs2{ii} = g(1)*s(ll).spdtfs(:,id1,:) + g(2)*s(ll).spdtfs(:,id2,:);
[s(ll).p2{ii},rang] = baumgartner2014(...
s(ll).dtfs2{ii},s(ll).spdtfs,s(ll).fs,'S',s(ll).S,...
'mrsmsp',MRS,'polsamp',polang,'rangsamp',1); % phantom
% total polar range
m2 = baumgartner2014_virtualexp(s(ll).p2{ii},pol0,rang,'runs',runs);
% R2 via correlation coefficient
r = corrcoef(m2(:,8),m2(:,6));
r2total(ll,idl) = r(2);
% restricted to frontal polar range
idfront = rang <= 90;
respangfront = rang(idfront);
idpol0front = pol0 <= 90;
pol0front = pol0(idpol0front);
p = s(ll).p2{ii}(idfront,idpol0front);
m2front = baumgartner2014_virtualexp(p,pol0front,respangfront,'runs',runs);
r = corrcoef(m2front(:,8),m2front(:,6));
r2front(ll,idl) = r(2);
% restricted to rear polar range
idrear = rang >= 90;
respangrear = rang(idrear);
idpol0rear = pol0 >= 90;
pol0rear = pol0(idpol0rear);
p = s(ll).p2{ii}(idrear,idpol0rear);
m2rear = baumgartner2014_virtualexp(p,pol0rear,respangrear,'runs',runs);
r = corrcoef(m2rear(:,8),m2rear(:,6));
r2rear(ll,idl) = r(2);
end
end
r2.total(:,ii) = nanmean(r2total,2);
r2.front(:,ii) = nanmean(r2front,2);
r2.rear(:,ii) = nanmean(r2rear,2);
end
amt_cache('set','loudspeakerspan_r2',r2,dPol)
end
% data extracted from Fig.6 (data:AVG) of Bremen et al. (2010, J Neurosci,
% 30:194-204) via http://arohatgi.info/WebPlotDigitizer
bremen2010.pol = 15:15:105;
bremen2010.r2 = [.85 , .81 , .63 , .38 , .19 , .11 , .21];
if flags.do_plot
figure
h(1) = plot(bremen2010.pol,bremen2010.r2,'bo-');
hold on
h(2) = plot(dPol,mean(r2.front),'bo-');
h(3) = plot(dPol,mean(r2.rear),'rs-');
h(4) = plot(dPol,mean(r2.total),'kd-');
set(h(1),'MarkerSize',kv.MarkerSize,'MarkerFaceColor','b')
set(h(2:4),'MarkerSize',kv.MarkerSize,'MarkerFaceColor','w')
set(gca,'XLim',[-5,110],'YLim',[-.05,1.05])
leg = legend('Frontal from [18]','Frontal','Rear','Overall','Location','best');
set(leg,'FontSize',kv.FontSize)
xlabel({'Loudspeaker span (deg)';''})
ylabel('\it{r}^{ 2}')
end
end
%% ------ FIG 9 of baumgartner2015jaes ------------------------------------
if flags.do_fig9
SysName{1} = 'NHK 22.2 (without bottom layer)';
LSPsetup{1} = [ 0,0 ; 30,0 ; 60,0 ; 90,0 ; 135,0 ; ...
180,0 ;-30,0 ;-60,0 ; -90,0 ;-135,0 ; ...
0,45; 45,45; 90,45; 135,45; 180,45; ...
-135,45;-90,45;-45,45; 0,90];
SysName{2} = 'Samsung 11.2';
LSPsetup{2} = [ 0,0 ; 60,0 ; 90,0 ; 135,0 ; ...
-60,0 ; -90,0 ;-135,0 ; ...
45,45;135,45; -45,45;-135,45];
SysName{3} = 'Samsung 10.2';
LSPsetup{3} = [ 0,0 ; 60,0 ; 90,0 ; 135,0 ; ...
-60,0 ; -90,0 ;-135,0 ; ...
45,45;180,45; -45,45];
SysName{4} = 'USC 10.2';
LSPsetup{4} = [ 0,0 ; 30,0 ; 60,0 ; 115,0 ; ...
180,0 ;-30,0 ;-60,0 ; -115,0 ; ...
45,45;-45,45];
SysName{5} = 'Auro-3D 10.1';
LSPsetup{5} = [ 30,0 ; 30,30 ; 135,30 ; 135,0;...
0,0 ;-30,0 ;-30,30 ;-135,30 ;-135,0; 0,90];
SysName{6} = 'Auro-3D 9.1';
LSPsetup{6} = [ 30,0 ; 30,30 ; 135,30 ; 135,0;...
0,0 ;-30,0 ;-30,30 ;-135,30 ;-135,0];
latall = -45:5:45;
polall = 0:10:180;
pe = amt_cache('get','locaVBAP',flags.cachemode);
if isempty(pe)
MRS = 0;
s = data_baumgartner2014('pool');
pe = zeros(length(latall),length(polall),length(s),length(LSPsetup),2); % predicted local polar RMS errors
for ll = 1:length(LSPsetup)
for aa = 1:length(latall)
lat = latall(aa);
for pp = 1:length(polall)
pol = polall(pp);
% Select LSP-Triangle and compute VBAP gains
[source_pos(1),source_pos(2)] = hor2sph(lat,pol);
Nlsp = length(LSPsetup{ll});
[g,IDsp] = local_vbap(LSPsetup{ll},source_pos);
for jj = 1:length(s)
% Compute binaural impulse response of loudspeaker triple
dtfs = permute(double(s(jj).Obj.Data.IR),[3 1 2]);
lsp_hrirs = zeros(length(IDsp),size(dtfs,1),2);
for ii = 1:length(IDsp)
% [lat_sp,pol_sp] = sph2horpolar(LSPsetup{ll}(IDsp(ii),1),LSPsetup{ll}(IDsp(ii),2));
idx = local_findNearestPoslocaVBAP(...
LSPsetup{ll}(IDsp(ii),1:2),s(jj).Obj.SourcePosition(:,1:2));
lsp_hrirs(ii,:,:) = squeeze(dtfs(:,idx,:));
end
target(:,1,1) = g*lsp_hrirs(:,:,1);
target(:,1,2) = g*lsp_hrirs(:,:,2);
% SP-template
[spdtfs,polang] = extractsp(lat,s(jj).Obj);
% Run loca model
[p,rang] = baumgartner2014(...
target,spdtfs,s(jj).fs,'S',s(jj).S,...
'lat',lat,'polsamp',polang,'mrsmsp',MRS);
m = baumgartner2014_virtualexp(p,pol,rang,'runs',1000);
pe(aa,pp,jj,ll,1) = localizationerror(m,'precPnoquerr');
[~,pe(aa,pp,jj,ll,2)] = baumgartner2014_pmv2ppp(p,pol,rang);
end
end
end
amt_disp([num2str(ll) ' of ' num2str(length(LSPsetup)) ' done']);
end
amt_cache('set','locaVBAP',pe)
end
MRS = 0;
pe_ref = amt_cache('get','locaVBAP_ref',flags.cachemode);
if isempty(pe_ref)
s = data_baumgartner2014('pool');
latall = -45:5:45;
polall = 0:10:180;
pe_ref = zeros(length(latall),length(polall),length(s),1,2); % predicted local polar RMS errors
%% Computations
for jj = 1:length(s)
dtfs = permute(double(s(jj).Obj.Data.IR),[3 1 2]);
for aa = 1:length(latall)
lat = latall(aa);
for pp = 1:length(polall)
pol = polall(pp);
[lat_sp,pol_sp,idx] = local_findNearestPoslocaVBAPref(lat,pol,s(jj).Obj.SourcePosition(:,1:2));
target = dtfs(:,idx,:);
% SP-template
[spdtfs,polang] = extractsp(lat,s(jj).Obj);
% Run loca model
[p,rang] = baumgartner2014(...
target,spdtfs,s(jj).fs,'S',s(jj).S,...
'mrsmsp',MRS,'lat',lat,'polsamp',polang);
m = baumgartner2014_virtualexp(p,pol,rang,'runs',1000);
pe_ref(aa,pp,jj,1,1) = localizationerror(m,'precPnoquerr');
[~,pe_ref(aa,pp,jj,1,2)] = baumgartner2014_pmv2ppp(p,pol,rang);
end
end
end
amt_cache('set','locaVBAP_ref',pe_ref)
end
N = length(LSPsetup);
eRMS = pe_ref(:,:,:,:,2);
eRMS(:,:,:,2:N+1) = pe(:,:,:,:,2);
if flags.do_plot
labels = {'Reference';'\it A';'\it B';'\it C';'\it D';'\it E';'\it F'};
labels = labels(1:N+1,:);
figure
for ll = 1:N+1
pemean = squeeze(mean(eRMS(:,:,:,ll),3));
subplot(1,N+2,ll)
imagesc(latall,polall,pemean')
set(gca,'YTick',0:30:180,'XTick',-30:30:30)
axis equal tight
colormap hot
if MRS == 0
ymin = 15.5;
ymax = 49.5;
else
ymin = 15.5;
ymax = 49.5;
end
caxis([ymin ymax])
if ll==1;
ylabel('Polar angle (deg)','FontName','Helvetica');
else
set(gca,'YTickLabel',[])
end
if ll==4; xlabel('Lateral angle (deg)','FontName','Helvetica'); end
title(labels{ll},'FontName','Helvetica')
% Loudspeaker positions
if ll > 1
[lat_lsp,pol_lsp] = sph2hor(LSPsetup{ll-1}(:,1),LSPsetup{ll-1}(:,2));
idlat = abs(lat_lsp) <= max(abs(latall))+1;
hold on
h2 = plot(lat_lsp(idlat),pol_lsp(idlat),'wo');
set(h2,'MarkerSize',2*3.5);
h1 = plot(lat_lsp(idlat),pol_lsp(idlat),'ko');
set(h1,'MarkerSize',2*3);
end
end
subplot(1,N+2,N+2)
y = ymin:1:ymax;
Ny = length(y);
pcolor(1:2,y,repmat(y(:),1,2))
shading flat
axis tight
colormap hot
title({' ';' '})
set(gca,'XTick',[],'YTick',20:5:45,... %,'TickLength',[0.12,0.12]
'YDir','normal','YAxisLocation','right','FontSize',kv.FontSize)
ylabel({'Polar error (deg)'},'FontSize',kv.FontSize)
set(gca,'Position',get(gca,'Position').*[1.03,1.8,0.3,0.8])
end
end
%% ------ Tab 1 of baumgartner2015jaes ------------------------------------
if flags.do_tab1
results = amt_cache('get','replicatePulkki2001',flags.cachemode);
[panang_varStrat,nCM_varStrat,p_varStrat,muhat,sigmahat] = amt_cache('get','replicatePulkki2001_varStrat',flags.cachemode);
if isempty(panang_varStrat)
exp_baumgartner2015('fig6','no_plot',flags.cachemode)
end
pulkki01 = data_pulkki2001;
[h1_pul,p1_pul] = kstest((pulkki01(1,:)-muhat(1))/sigmahat(1)); % center data acc. to target distribution and then test similarity to standard normal distribution
[h2_pul,p2_pul] = kstest((pulkki01(2,:)-muhat(2))/sigmahat(2));
[h1_pm,p1_pm] = kstest((results.panang_Pmax(1,:)-muhat(1))/sigmahat(1)); % center data acc. to target distribution and then test similarity to standard normal distribution
[h2_pm,p2_pm] = kstest((results.panang_Pmax(2,:)-muhat(2))/sigmahat(2));
[h1_cm,p1_cm] = kstest((results.panang_Cen(1,:)-muhat(1))/sigmahat(1)); % center data acc. to target distribution and then test similarity to standard normal distribution
[h2_cm,p2_cm] = kstest((results.panang_Cen(2,:)-muhat(2))/sigmahat(2));
amt_disp('p-values of K.S.-test:','documentation');
amt_disp('Real source at 0 deg:','documentation');
amt_disp([' Results Pulkki (2001): p = ' num2str(p1_pul,'%3.2f')],'documentation');
amt_disp([' Probability Maximiz.: p = ' num2str(p1_pm,'%3.2f')],'documentation');
amt_disp([' Centroid Match: p = ' num2str(p1_cm,'%3.2f')],'documentation');
amt_disp([' Individual strategy: p = ' num2str(p_varStrat(1),'%3.2f') ' (#CM = ' num2str(nCM_varStrat) ')'],'documentation');
amt_disp('Real source at 15 deg:','documentation');
amt_disp([' Results Pulkki (2001): p = ' num2str(p2_pul,'%3.2f')],'documentation');
amt_disp([' Probability Maximi.: p = ' num2str(p2_pm,'%3.2f')],'documentation');
amt_disp([' Centroid Match: p = ' num2str(p2_cm,'%3.2f')],'documentation');
amt_disp([' Individual strategy: p = ' num2str(p_varStrat(2),'%3.2f') ' (#CM = ' num2str(nCM_varStrat) ')'],'documentation');
end
%% ------ Tab 3 of baumgartner2015jaes ------------------------------------
if flags.do_tab3
pe = amt_cache('get','locaVBAP',flags.cachemode);
pe_ref = amt_cache('get','locaVBAP_ref',flags.cachemode);
if isempty(pe)
exp_baumgartner2014('fig9_baumgartner2015jaes','no_plot',flags.cachemode);
end
N = size(pe,4); % # loudspeakers
eRMS = pe_ref(:,:,:,:,2);
eRMS(:,:,:,2:N+1) = pe(:,:,:,:,2);
labels = {'Reference';'\it A';'\it B';'\it C';'\it D';'\it E';'\it F'};
labels = labels(1:N+1,:);
amt_disp('RMS error difference from reference averaged across directions','documentation');
amt_disp('System min mean max','documentation');
for ll = 2:N+1
IeRMS = eRMS(:,:,:,ll) - eRMS(:,:,:,1);
IeRMS = mean(IeRMS,3); % average across listeners
amt_disp([labels{ll} ' ' num2str(min(IeRMS(:)),'%2.1f') ' ' num2str(mean(IeRMS(:)),'%2.1f') ' ' num2str(max(IeRMS(:)),'%2.1f')],'documentation');
end
end
end
%% ------------------------------------------------------------------------
% ---- INTERNAL FUNCTIONS ------------------------------------------------
% ------------------------------------------------------------------------
function [idx,posN] = local_findNearestPoslocaVBAP(posdesired,posexist)
% FINDNEARESTPOS_LOCAVABAP finds nearest position. Data given in spherical
% coordinates
%
% Usage: [idx,posN] = findNearestPos(posdesired,posexist)
ds = deg2rad(posdesired);
es = deg2rad(posexist);
[d(:,1),d(:,2),d(:,3)] = sph2cart(ds(:,1),ds(:,2),ones(size(ds,1),1));
[e(:,1),e(:,2),e(:,3)] = sph2cart(es(:,1),es(:,2),ones(size(es,1),1));
D = e-repmat(d,length(es),1);
[Dmin,idx] = min(sum(D.^2,2));
posN = posexist(idx,:);
end
function [latN,polN,idx] = local_findNearestPoslocaVBAPref(lat,pol,aziele)
% FINDNEARESTPOS_LOCAVBAP_REF finds nearest position according to lat. and pol. angle
%
% Usage: [latN,polN,idx] = findNearestPos(lat,pol,aziele)
[positions(:,1),positions(:,2)] = sph2hor(aziele(:,1),aziele(:,2));
d_lat = abs(lat-positions(:,1));
d_pol = abs(pol-positions(:,2));
d = d_lat+d_pol;
[d_min,idx] = min(d);
latN = positions(idx,1);
polN = positions(idx,2);
end
function [g,IDspeaker] = local_vbap(lsp_coord,source_pos)
%VBAP Returns array of gains for VBAP triplet of speakers.
% Usage: [g,IDspeaker] = vbap(speaker_coord,source_pos)
%
% Input Parameters:
% lsp_coord : spherical (azi,ele) coordinates of loudspeakers
% source_pos : spherical (azi,ele) coordinates of phantom source
%
% Output Parameters:
% g : VBAP gains of loudspeaker triplet
% IDspeaker : indices of selected loudspeaker triplet
Nlsp = size(lsp_coord,1);
% Convert to spherical coordinates winth angles in radians
speaker_sph = deg2rad(lsp_coord);
source_sph = deg2rad(source_pos);
% Convert to cartesian coordinates
[speaker_cart(:,1),speaker_cart(:,2),speaker_cart(:,3)] = sph2cart(...
speaker_sph(:,1),speaker_sph(:,2),ones(Nlsp,1));
[source_cart(1),source_cart(2),source_cart(3)] = sph2cart(...
source_sph(1),source_sph(2),1);
% Add imaginary speaker below
if min(lsp_coord(:,2)) >= 0
speaker_cart = [speaker_cart;0,0,-10];
end
% Select lsp. triplet
dt = DelaunayTri(speaker_cart);
ch = convexHull(dt);
% figure; trisurf(ch, dt.X(:,1),dt.X(:,2),dt.X(:,3), 'FaceColor', 'cyan')
d = zeros(length(ch),1);
for ii = 1:length(ch)
d(ii) = sum(local_dist(source_cart,speaker_cart(ch(ii,:),:)));
end
[~,IDch] = min(d);
IDspeaker = ch(IDch,:);
% Compute lsp. gains
g = source_cart / speaker_cart(IDspeaker,:);
g = g / norm(g);
end
function d = local_dist(x,Y)
d = sqrt(sum((repmat(x,size(Y,1),1)-Y).^2,2));
end
Build with Bootstrap