function msq = exp_steidle2019(varargin)
%EXP_STEIDLE2019 Compares various ITD calculation methods
%   Usage: exp_steidle2019(flag) 
%
% 
%   The following flags can be chosen:
%
%     'clean'       clean data
%
%     'noisy'       noisy data
%
%     'stim'        to create the data which was used
%
%     'figure'      to create the plots shown in the paper (fig1,...fig9)
%
%     'lowpass'     (optional) Decide if lowpass shall be applied. 
%                   lp for lowpass (default), bb for broadband
%
%     'threshlvl'   (optional) Set threshold level for 'Threshold' mode in dB.        
%                   Default is -10 dB. 
%
% 
% 
%   Requirements: 
%   -------------
%
%   1) SOFA API from http://sourceforge.net/projects/sofacoustics for Matlab (in e.g. thirdparty/SOFA)
% 
%
%   Examples:
%   ---------
%
%   To display results for Fig.1 use :
%
%     exp_steidle2019('fig1');
% 
%   To display results for Fig.2 use :
%
%     exp_steidle2019('fig2');
%  
%   To recalculate the data sets use i.e. :
%
%     exp_steidle2019('clean','lp');
% 
%      
%   References:
%     L. Steidle and R. Baumgartner. Geometrical evaluation of methods to
%     approximate interaural time differences by broadband delays. In
%     Proceedings of the German Annual Meeting (DAGA), Rostock, DE, Mar 2019.
%     
%
%   Url: http://amtoolbox.org/amt-1.4.0/doc/experiments/exp_steidle2019.php


%   #Author: Laurin Steidle

% 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. 

    
definput.import={'amt_cache'}; % get the flags of amt_cache
definput.flags.choose = {'clean','noisy','stim',...
    'fig1','fig2','fig3','fig4','fig5','fig6','fig7','fig8','fig9'};
definput.flags.lp = {'lp','bb'};
definput.keyvals.threshlvl = -10;
definput.keyvals.butterpoly = 6;

[flags,kv]=ltfatarghelper({},definput,varargin);

%% ---------------calculate-estimation-data-------------------------

% -------------------------------------------------------------------------
% calulates ITDs with all models for 'clean' synthesised IRs    
    if flags.do_clean

        modes = {'Threshold','Cen_e2','MaxIACCr', 'MaxIACCe',...
                'CenIACCr', 'CenIACCe', 'CenIACC2e', 'PhminXcor','IRGD'};
              
        n_modes = size(modes,2);

        
        % importing a set of head parameter from ziegelwanger2014
        tmp=amt_load('ziegelwanger2014','info.mat');
        data=tmp.info.Rotation;  
        data.radius = data.radius*1e-3;     % head radius in meter
        data.phi = data.phi*pi/180;         % polar angel in radian
        data.theta  = data.theta*pi/180;    % azimut angel in radian

        head_idx = 15:15+13;        % chosing data subset
        n_heads = size(head_idx,2); 
        
        % calculating theoretical and estimated itd's for all head param
        % ## initiate array
        msq = zeros(n_heads,n_modes); 
        
        for ii=head_idx % goes thru all head geometries
            Obj=amt_load('ziegelwanger2014',['Sphere_Rotation_' data.subjects{ii} '.sofa']);  
            
            %calculate theoretically expected ITDs
            p0_offaxis = [[data.radius(ii);0; 0; 0; 0;data.phi(ii)+pi/2;  data.theta(ii)*pi/180]...
                          [data.radius(ii);0; 0; 0; 0;data.phi(ii)-pi/2; -data.theta(ii)*pi/180]];
            toa_sim_left    = ziegelwanger2014_offaxis(p0_offaxis(:,1),Obj.SourcePosition(:,1:2)*pi/180);
            toa_sim_right   = ziegelwanger2014_offaxis(p0_offaxis(:,2),Obj.SourcePosition(:,1:2)*pi/180);
            toadiff_sim     = toa_sim_left - toa_sim_right;  
            
            for kk = 1:n_modes % goes thru all estimation methods
            %calculate estimated ITDs
                toadiff = itdestimator(Obj,modes{kk}, flags.lp, 'butterpoly', kv.butterpoly); 
                msq(ii-n_heads,kk) = sqrt(sum( (toadiff-toadiff_sim).^2 )/size(toadiff,1));         
            end
        end
    end



% -------------------------------------------------------------------------
    if flags.do_noisy
        
        modes = {'Threshold','Cen_e2','MaxIACCr', 'MaxIACCe',...
                'CenIACCr', 'CenIACCe', 'CenIACC2e', 'PhminXcor','IRGD'};
        n_modes = size(modes,2);
        
        % importing a set of head parameter from ziegelwanger2014
        tmp=amt_load('ziegelwanger2014','info.mat');
        data=tmp.info.Rotation;  
        data.radius = data.radius*1e-3;     % head radius in meter
        data.phi = data.phi*pi/180;         % polar angel in radian
        data.theta  = data.theta*pi/180;    % azimut angel in radian

        head_idx = 15:15+13;        % chosing data subset
        n_heads = size(head_idx,2);
                        
        %define added noise levels of noise in dB of SNR
        noise_lvl = [50,40,30,20];
                
        % calculating theoretical and estimated itd's for all head param
        % ## initiate array               
        msq = zeros(n_heads,size(noise_lvl,2),n_modes); 
        
        for ii=head_idx % goes thru all head geometries
            for jj = 1:size(noise_lvl,2) % level of noise
                Obj=amt_load('ziegelwanger2014',['Sphere_Rotation_' data.subjects{ii} '.sofa']);

                % calculate theoretically expected ITDs
                p0_offaxis = [[data.radius(ii);0; 0; 0; 0;data.phi(ii)+pi/2;  data.theta(ii)*pi/180]...
                              [data.radius(ii);0; 0; 0; 0;data.phi(ii)-pi/2; -data.theta(ii)*pi/180]];
                toa_sim_left    = ziegelwanger2014_offaxis(p0_offaxis(:,1),Obj.SourcePosition(:,1:2)*pi/180);
                toa_sim_right   = ziegelwanger2014_offaxis(p0_offaxis(:,2),Obj.SourcePosition(:,1:2)*pi/180);
                toadiff_sim     = toa_sim_left - toa_sim_right;  
                
                % adds noise
                for kk=1:Obj.API.M
                    for mm=1:Obj.API.R
                        ir_tmp = squeeze(Obj.Data.IR(kk,mm,:));
                        Obj.Data.IR(kk,mm,:) = ir_tmp + ...
                          scaletodbspl(noise(length(ir_tmp),1),dbspl(ir_tmp)-noise_lvl(jj),100);
                    end
                end

                for kk = 1:n_modes % goes thru all estimation methods
                %calculate estimated ITDs
                    toadiff = itdestimator(Obj,modes{kk},flags.lp,'threshlvl',kv.threshlvl, 'butterpoly', kv.butterpoly);
                    msq(ii-n_heads,jj,kk) = sqrt(sum( (toadiff-toadiff_sim).^2 )/size(toadiff,1));      
                end
            end
        end
    end

    
% -------------------------------------------------------------------------
% creates data based on stimuli
% this is done to compare models found in itdestimator to the dietz model

    if flags.do_stim
        
        modes = {'Dietz'};
        %modes = {'MaxIACCr', 'MaxIACCe','CenIACCr', 'CenIACCe', 'CenIACC2e','IRGD','Dietz'};
        
        n_modes = size(modes,2);
        
         % importing a set of head parameter from ziegelwanger2014
        tmp=amt_load('ziegelwanger2014','info.mat');
        data=tmp.info.Rotation;  
        data.radius = data.radius*1e-3;     % head radius in meter
        data.phi = data.phi*pi/180;         % polar angel in radian
        data.theta  = data.theta*pi/180;    % azimut angel in radian

        head_idx = 15:15+13;        % chosing data subset
        n_heads = size(head_idx,2);
        
        % calculating theoretical and estimated itd's for all head param
        % ## initiate array                    
        msq = zeros(n_heads,n_modes);

        for ii=head_idx
            Obj=amt_load('ziegelwanger2014',[ 'Sphere_Rotation_' data.subjects{ii} '.sofa']);
            % creates white noise and adjusts object samlpe length
            fs  = Obj.Data.SamplingRate;
            white_noise = noise(fs/2);
            Obj.API.N = size(Obj.Data.IR,3) + size(white_noise,1) - 1;
       
            % adds white noise
            stim = zeros(Obj.API.M, Obj.API.R, Obj.API.N);
            for kk=1:Obj.API.M      %pos
                for mm=1:Obj.API.R  %ear
                    stim(kk,mm,:) = lconv(Obj.Data.IR(kk,mm,:),white_noise);
                end
            end
            Obj.Data.IR = stim;

            % calculate theoretically expected ITDs
            p0_offaxis = [[data.radius(ii);0; 0; 0; 0;data.phi(ii)+pi/2;  data.theta(ii)*pi/180]...
                          [data.radius(ii);0; 0; 0; 0;data.phi(ii)-pi/2; -data.theta(ii)*pi/180]];
            toa_sim_left    = ziegelwanger2014_offaxis(p0_offaxis(:,1),Obj.SourcePosition(:,1:2)*pi/180);
            toa_sim_right   = ziegelwanger2014_offaxis(p0_offaxis(:,2),Obj.SourcePosition(:,1:2)*pi/180);
            toadiff_sim     = toa_sim_left - toa_sim_right;        

            % calculate estimated ITDs
            for kk = 1:n_modes    
               if strcmp(modes{kk},'Dietz')
               fprintf('Dietz \n')
                    for nn=1:Obj.API.M
                       insig = transpose(squeeze(stim(nn,:,:)));
                       dietz_out = dietz2011(insig,fs);
                       if flags.do_lp
                        med = median(dietz_out.itd_lp);
                       else
                        med = median(dietz_out.itd);
                       end
                       % Care: Dietz always uses a FIXED frequency range!
                       toadiff(nn) = mean(med(1:8)); % <-- { med(1:8) }
                       toadiff = transpose(toadiff);
                    end
                else
                   toadiff = itdestimator(Obj,modes{kk},flags.lp, 'butterpoly', kv.butterpoly);
                end              
                msq(ii,kk) = sqrt(sum( (toadiff-toadiff_sim).^2 )/size(toadiff,1));
            end         
        end
    end
    
%% ---------------figure-1------------------------------------------
% Visualisation of Threshold construction examplary for a synthesized
% binaural HRIR of a spherical head with a sound source positioned at an 
% azimut angle of 20 

    if flags.do_fig1
        nn = 115;
        col = jet(5);
        Obj = amt_load('baumgartner2017','hrtf b_nh14.sofa');
        fs = Obj.Data.SamplingRate;
        
        fig = figure;

        [a1,~] = findpeaks(squeeze(Obj.Data.IR(nn,1,:)),'MinPeakProminence',0.005);
        th_value1 = max(a1)*0.2;
        [x1,~] = find(squeeze(Obj.Data.IR(nn,1,:))>th_value1,1);

        [a2,~] = findpeaks(squeeze(Obj.Data.IR(nn,2,:)),'MinPeakProminence',0.005);
        th_value2 = max(a2)*0.2;
        [x2,~] = find(squeeze(Obj.Data.IR(nn,2,:))>th_value2,1);

        ax1 = subplot(2,1,1);
        hold on
        plot((0:size(squeeze(Obj.Data.IR(nn,1,:)),1)-1)/fs*1e3,...
            squeeze(Obj.Data.IR(nn,1,:)),'k-')
        plot(ax1,[x1/fs*1e3; x1/fs*1e3],[-1; 1],'Color',col(5,:))
        plot(ax1,[x2/fs*1e3; x2/fs*1e3],[-1; 1],'Color',col(3,:))
        plot(ax1,[0; size(Obj.Data.IR(nn,1,:),3)/fs*1e3],[th_value1; th_value1],'Color',col(5,:))
        hold off
        title(ax1,'Left ear')
        xlim(ax1,[0 3])
        set(ax1,'Ytick',[-0.03,0,0.03])
        xlabel('Time (ms)')
        ylim(ax1,[-0.045 0.045])
        ylabel('Amplitude (a.u)')

        ax2 = subplot(2,1,2); 
        hold on
        plot((0:size(squeeze(Obj.Data.IR(nn,2,:)),1)-1)/fs*1e3,...
            squeeze(Obj.Data.IR(nn,2,:)),'k-')
        plot(ax2,[x2/fs*1e3; x2/fs*1e3],[-1; 1],'Color',col(5,:))
        plot(ax2,[0; size(Obj.Data.IR(nn,2,:),3)/fs*1e3],[th_value2; th_value2],'Color',col(5,:))
        hold off
        title(ax2,'Right ear')
        xlim(ax2,[0 3])
        set(ax2,'Ytick',[-0.03,0,0.03])
        xlabel('Time (ms)')
        ylim(ax2,[-0.045 0.045])
        ylabel('Amplitude (a.u.)')
        

    end

%% ---------------figure-2------------------------------------------
% IACC of synthesized binaural HRIR of a spherical head with a sound
% source positioned at an azimut angle of 20 

     if flags.do_fig2

        tmp=amt_load('ziegelwanger2014','info.mat');
        data=tmp.info.Rotation;  
        data.radius = data.radius*1e-3;
        data.phi = data.phi*pi/180;
        data.theta  = data.theta*pi/180;       
%         n_heads = size(data.subjects,2);
                
        fig = figure();
        hold on
        for ii=3:3 %n_heads
            jj = 115;
            Obj=amt_load('ziegelwanger2014',[ 'Sphere_Rotation_' data.subjects{ii} '.sofa']);

            plot(transpose(-239:239)/Obj.Data.SamplingRate*1e3,...
                envelope(xcorr(squeeze(Obj.Data.IR(jj,1,:)),squeeze(Obj.Data.IR(jj,2,:)))))
            plot(transpose(-239:239)/Obj.Data.SamplingRate*1e3,...
                xcorr(squeeze(Obj.Data.IR(jj,1,:)),squeeze(Obj.Data.IR(jj,2,:))))
        end
        %title( 'Examplary inter aural cross correlation' )
        xlabel('time delay (ms)')
        ylabel('correlation (a.u.)')
        legend('envelope of cc','cross correlation','Location','best')
        xlim([-1 1])
        hold off
        
     end
     
%% ---------------figure-3------------------------------------------
% Group delay of synthesized binaural HRIR of a spherical head with a
% sound source positioned at an azimut. angle of 20 
    if flags.do_fig3
        

        tmp=amt_load('ziegelwanger2014','info.mat');
        data=tmp.info.Rotation;  
%         n_heads = size(data.subjects,2);
                
        fig = figure();
        hold on
        for ii=5:5 %n_heads
            jj = 115;
            Obj=amt_load('ziegelwanger2014',['Sphere_Rotation_' data.subjects{ii} '.sofa']);
            Ns  = Obj.API.N;
            fs  = Obj.Data.SamplingRate;
            [gd,w] = grpdelay(squeeze( Obj.Data.IR(jj,1,:) ),1,Ns,fs );
            plot(w,gd/fs*1e3)
            
        end
        title( 'Exemplary group delay' )
        xlabel('Frequency (Hz)')
        ylabel('Group delay (ms)')
        xlim([500,5000])
        ylim([1.35 1.8])
        hold off
        
    end

%% ---------------figure-4------------------------------------------
% -------------------------------------------------------------------------
% ITD estimations along horizontal plane for listener for NH15

    if flags.do_fig4

        Obj = amt_load('baumgartner2017','hrtf b_nh15.sofa');
        white_noise = noise(258);

        % ---------------------- calculating toa ----------------------------------

        modes = {'Threshold','Cen_e2','MaxIACCr', 'MaxIACCe',...
                'CenIACCr', 'CenIACCe', 'CenIACC2e', 'PhminXcor','IRGD','Dietz'};
        n_modes = size(modes,2);
        cc = hsv(n_modes);

        plane_idx = find( Obj.SourcePosition(:,2) == 0 );
        plane_angle = Obj.SourcePosition(plane_idx,1);
        [n_angles,~] = size(plane_angle);

        for kk=1:n_modes
            if strcmp(modes{kk},'Dietz')
                fprintf('Dietz \n')
                for nn=1:n_angles
                     stimulus = SOFAspat(white_noise,Obj,plane_angle(nn),0);
                     dietz_out = dietz2011(stimulus,Obj.Data.SamplingRate);
                     med = median(dietz_out.itd);
                     toa_diff{kk}(nn) = mean(med(1:8));
                end
            else
            toa_diff{kk} = itdestimator(Obj,modes{kk}, 'butterpoly', kv.butterpoly);  
            end
        end
    
        fig = figure();
        hold on
        for kk=1:n_modes
            if strcmp(modes{kk},'Dietz') || strcmp(modes{kk},'Dietz lp')
                plot(plane_angle,toa_diff{kk},'color', cc(kk,:))
            else
                plot(plane_angle,toa_diff{kk}(plane_idx),'color', cc(kk,:))    
            end
        end
        
        xlabel('azimuthal angle (Deg)')
        ylabel('interaural time difference (s)')
        xlim([0,360])
        legend('Threshold','Cen_e2','MaxIACCr', 'MaxIACCe',...
                'CenIACCr', 'CenIACCe', 'CenIACC2e', 'PhminXcor',...
                'IRGD','Dietz','Location','eastoutside') 
        hold off
        
    end
%% ---------------figure-5------------------------------------------
% -------------------------------------------------------------------------
% Ear positions of all heads. phi represents the azimuth angle while theta
% shows elevations.
    
    if flags.do_fig5

    %load data
        data=data_ziegelwanger2014('SPHERE_ROT',flags.cachemode);
        for ii=1:length(data.results)
            p_onaxis{1}(:,:,ii)=data.results(ii).MAX{1}.p_onaxis;
            p_onaxis{2}(:,:,ii)=data.results(ii).CTD{1}.p_onaxis;
            p_onaxis{3}(:,:,ii)=data.results(ii).AGD{1}.p_onaxis;
            p_onaxis{4}(:,:,ii)=data.results(ii).MCM{1}.p_onaxis;
        end

    %plot figure
        fig = figure();

        lw2=2;  %linewidth
        ls='kx';%linestyle
        lc=[0 0 0];
        
        %phi
        subplot(211);
        var=[squeeze(p_onaxis{4}(2,1,:))/pi*180 ...
            squeeze(p_onaxis{4}(2,2,:))/pi*180 ...
            data.phi+ones(length(data.phi),1)*90 ...
            data.phi-ones(length(data.phi),1)*90];

        hold on
        for ch=1:size(p_onaxis{4},2)
            plot(15:23,var(15:23,2+ch),ls,'Linewidth',lw2,'color',lc)
            plot(24:28,var(24:28,2+ch),ls,'Linewidth',lw2,'color',lc)
        end

        clear var;
        set(gca,'YTick',[-90 90])
        set(gca,'xtick',1:42)
        xlim([15-1,15+14])
        set(gca,'Xticklabel',{''})
        ylabel('\phi (deg)')
        grid on
        set(gca,'GridLineStyle','--')
        
        
        %theta
        subplot(212);
        var=[squeeze(p_onaxis{4}(3,1,:))/pi*180 ...
            squeeze(p_onaxis{4}(3,2,:))/pi*180 ...
            data.theta ...
            -data.theta];

        hold on
        for ch=1:size(p_onaxis{4},2)
            plot(15:23,var(15:23,2+ch),ls,'Linewidth',lw2,'color',lc)
            plot(24:28,var(24:28,2+ch),ls,'Linewidth',lw2,'color',lc)           
        end

        clear var;
        ylabel('\theta (deg)')
        xlabel('Condition')
        ylim([-15,15])
        xlim([15-1,15+14])
        set(gca,'xtick',1:42)
        set(gca,'xticklabel',[])
        set(gca,'ytick',[-10 0 10])
        grid on
        set(gca,'GridLineStyle','--')
        
        %set(gcf,'color',[1 1 1]) 
    end


%% ---------------figure-6------------------------------------------
% -------------------------------------------------------------------------    
% Synthesized binaural HRIR of a spherical head with a sound source
% positioned at an azimuth. angle of 20 with additive noise

    if flags.do_fig6

        tmp=amt_load('ziegelwanger2014','info.mat');
        data=tmp.info.Rotation;  
        data.radius = data.radius*1e-3;
        data.phi = data.phi*pi/180;
        data.theta  = data.theta*pi/180;

        noise_lvl = [20,30,40];

        n_noise = size(noise_lvl,2);
        
        
        fig = figure();
        hold on
        for jj = 1:n_noise % level of noise
            Obj=amt_load('ziegelwanger2014',[ 'Sphere_Rotation_' data.subjects{4} '.sofa']);

            IR = squeeze(Obj.Data.IR(777,1,:));
            IR = IR + scaletodbspl(noise(length(IR),1),dbspl(IR)-noise_lvl(jj), 100);%%awgn(IR,noise_lvl(jj));

            plot(IR)

        end
        %title( 'Impulse response with added gaussian noise' )
        xlabel('Sample point')
        xlim([0 240])
        ylabel('Amplitude (a.u.)')
        legend('SNR 20dB','SNR 30dB','SNR 40dB','Location','best')
        hold off
        
    end
    
    
%% ---------------figure-7------------------------------------------
% ANR for all examined heads exemplary for the threshold estimation method.

    if flags.do_fig7
        msq = amt_cache('get','clean_bb_msq',flags.cachemode);
        if isempty(msq)
          msq = exp_steidle2019('clean', 'bb');
          amt_cache('set','clean_bb_msq',msq);
        end
        msq_bb_clean = msq;
        
        fig = figure();
        bar(msq_bb_clean(:,1,1)*1e6)
        xlabel('Heads as described in section 3.1')
        ylabel('ANR in \mus')
        colormap('lines')
        set(gca, 'YMinorGrid','on', 'YMinorGrid','on')
        
    end
    
    
%% ---------------figure-8------------------------------------------
% ANR for all examined estimation methods and SNRs, broadband &
% low-pass (shown in black)

    if flags.do_fig8
        
        % msq_bb_clean
        msq_bb_clean = amt_cache('get','clean_bb_msq',flags.cachemode);
        if isempty(msq_bb_clean)
          msq_bb_clean = exp_steidle2019('clean', 'bb');
          amt_cache('set','clean_bb_msq',msq_bb_clean);
        end
        
        % msq_lp_clean
        msq_lp_clean = amt_cache('get','clean_lp_msq',flags.cachemode);
        if isempty(msq_lp_clean)
          msq_lp_clean = exp_steidle2019('clean', 'lp');
          amt_cache('set','clean_lp_msq',msq_lp_clean);
        end

        % msq_bb_noisy
        msq_bb_noisy = amt_cache('get','noisy_bb_msq',flags.cachemode);
        if isempty(msq_bb_noisy)
          msq_bb_noisy  = exp_steidle2019('noisy', 'bb');
          amt_cache('set','noisy_bb_msq',msq_bb_noisy);
        end

        % msq_lp_noisy
        msq_lp_noisy = amt_cache('get','noisy_lp_msq',flags.cachemode);
        if isempty(msq_lp_noisy)
          msq_lp_noisy  = exp_steidle2019('noisy', 'lp');
          amt_cache('set','noisy_lp_msq',msq_lp_noisy);
        end

        
        msq_bb_clean = reshape(msq_bb_clean,14,1,9);
        msq_bb = cat(2,msq_bb_clean,msq_bb_noisy);
        msq_bb = msq_bb(1:end-5,:,:);
%         msq_bb_std = squeeze( std(msq_bb) );
        msq_bb_sum = squeeze(sum(msq_bb,1)/size(msq_bb,1));
                
        msq_lp_clean = reshape(msq_lp_clean,14,1,9);
        msq_lp = cat(2,msq_lp_clean,msq_lp_noisy);
        msq_lp = msq_lp(1:end-5,:,:);
%         msq_lp_std = squeeze( std(msq_lp )); 
        msq_lp_sum = squeeze(sum(msq_lp,1)/size(msq_lp,1));
        
        
        msq_int_sum = reshape(cat(2,cat(1,cat(1,msq_bb_sum,msq_lp_sum),zeros(5,9)),zeros(15,1)),5,30);
%         msq_int_std = reshape(cat(2,cat(1,cat(1,msq_bb_std,msq_lp_std),zeros(5,9)),zeros(15,1)),5,30);
        M = size(msq_int_sum,1);
        N = size(msq_int_sum,2);
        K = numel(msq_int_sum);
        g = 0; k=0;

        bb_idx = 1:3:N-4;
        lp_idx = 2:3:N-4;
%         zo_idx = [3:3:N-4,28,29,30];
%         len = size(bb_idx,2);
        
        fig = figure();
        hold on
        col = parula(10);
        for ii=1:M
            for jj = 1:N
                if any(bb_idx == jj)
                    handle(mod(g,9)+1) = barh(K,msq_int_sum(ii,jj)*1e6,1,'FaceColor',col(mod(g,9)+1,:));
%                     errorbar(K,msq_int_sum(ii,jj),msq_int_std(ii,jj),...
%                         'MarkerEdgeColor','k','MarkerFaceColor','w')
                    g = g+1;
                elseif any(lp_idx == jj)
                    barh(K,msq_int_sum(ii,jj)*1e6,1,'FaceColor',[1,1,1]*0.2)%col(mod(k,9)+1,:),'EdgeColor','r')
%                     errorbar(K,msq_int_sum(ii,jj),msq_int_std(ii,jj),...
%                         'MarkerEdgeColor','k','MarkerFaceColor',[.49 1 .63])
                    k = k+1;
                end 
                K = K-1;
            end
        end
        xlim([5e0,2e3])
        xlabel('ANR in \mus')
        ylabel('SNR in dB')
        set(gca, 'XScale', 'log')
        set(gca, 'yTickLabel',{'20 dB','30 dB','40 dB','50 dB','clean'},'YTick',[18,48,78,108,138])
        set(gca, 'XMinorGrid','on', 'XMinorGrid','on')
               
        legend(handle, 'Threshold','Cen_e2','MaxIACCr', 'MaxIACCe',...
                    'CenIACCr', 'CenIACCe', 'CenIACC2e', 'PhminXcor','IRGD','Location','northeast')
        
       
    end
    
    
%% ---------------figure-9------------------------------------------
% ANR for chosen estimation methods (broadband & low-pass) based
% on binaural signals

    if flags.do_fig9
        
        % stim_bb
        msq_bb_stim = amt_cache('get','stim_bb_msq',flags.cachemode);
        if isempty(msq_bb_stim)
          msq_bb_stim = exp_steidle2019('stim', 'bb');
          amt_cache('set','stim_bb_msq',msq_bb_stim);
        end
        
        % stim_lp
        msq_lp_stim = amt_cache('get','stim_lp_msq',flags.cachemode);
        if isempty(msq_lp_stim)
          msq_lp_stim = exp_steidle2019('stim', 'lp');
          amt_cache('set','stim_lp_msq',msq_lp_stim);
        end
        
        msq_stim = cat(2,msq_bb_stim,msq_lp_stim);
        msq_stim_sum = squeeze(sum(msq_stim,1)/size(msq_stim,1));
%         msq_stim_std = std(msq_stim);
        
        
       
        fig = figure();
        hold on
        modes = {'MaxIACCr', 'MaxIACCe','CenIACCr', 'CenIACCe', 'CenIACC2e',...
                 'IRGD','Dietz',...
                 'MaxIACCr lp', 'MaxIACCe lp','CenIACCr lp', 'CenIACCe lp',...
                 'CenIACC2e lp','IRGD lp','Dietz lp'};
        barh(1:size(msq_stim_sum,2),msq_stim_sum*1e6)
%         errorbar(msq_stim_sum*1e6,1:size(msq_stim_sum,2),msq_stim_std*1e6,'horizontal','k.')
%                          'MarkerEdgeColor','k','MarkerFaceColor',[.49 1 .63])
        xlim([0,500])
        xlabel('ANR in \mus')
        
        set(gca, 'YTickLabel',modes, 'YTick',1:size(msq_stim_sum,2))
        colormap('lines')
        set(gca, 'XMinorGrid','on', 'XGrid','on')
        hold off
        
                
    end 
end