THE AUDITORY MODELING TOOLBOX
Go to function
HAUTH2020 - Blind equalization cancellation model
Program code:
function out_struct = hauth2020(MixedSig, fs, varargin)
%HAUTH2020 Blind equalization cancellation model
%
% Usage: out_struct = hauth2020('RequiredSignal',mixed_input,'OptionalSignal',...[OptionalSignal1 OptionalSignal2]...,OptionalSignalN,'model_params',model_params)
%
% Input parameters:
% MixedSig : two-channel matrix containing signal + noise
% fs : the sampling frequency of RequiredSignal [Hz]
%
% Output parameters:
% out_struct : Structure containing the processed signals, in the order [Sel Min Max]
%
% Additional input parameters:
%
% 'OptSigs', OptSig matrix containing one or more optional two channel signals
%
% 'fmin', fmin center frequency of the lowest filter of the gammatone filterbank [Hz]
%
% 'fmax', fmax center frequency of the highest filter of the gammatone filterbank [Hz]
%
% 'f_target' , f_target specified frequency that will have a matched gammatone filter [Hz]
%
% 'ERB_factor', ERB_factor gammatone filter bandwidth
%
% Hauth2020 accepts the following flags:
%
% 'binauralinaccuracies' enable binaural processing inaccuracies
%
% 'no_binauralinaccuracies' disable binaural processing inaccuracies
%
% 'longterm' SRMR-decision making considers whole signal
%
% 'shortterm' blockwise SRMR-decision making based on same time constants as EC and BE processing
%
%
% Parameter description:
%
% 'RequiredSignal' is the mixture of speech and noise, which is a two
% channel (left,right) matrix (mixed_input)
%
% 'OptionalSignal' One or more optional two channel signals, which are
% processed in the same way as the 'RequiredSignal'. They have to be
% arranged in a single matrix such that always two columns correspond to
% the left and right ear signals [left1 right1 left2 right2 ... leftN rightN]
% (e.g. if you need clean speech and noise for your back-end)
%
% 'model_params' Structure containing model parameters like frequency range etc.
%
% out_struct Structure containing the processed signals, in the order [Sel Min Max]
% e.g. out_struct.signals.Mixsigsynfin contains the EC/BE processed mixed
% signal, where the best strategy was chosen blindly
%
% The SII used in the study by Hauth et al. (2020)
% requires band specific levels. If optional signals are present,
% the levels of all optional signals after EC processing and better ear
% selection are part of the output structure.
%
% See also: demo_hauth2020 hauth2020_fftcon hauth2020_sii hauth2020_ecprocess4optsigs
% hauth2020_srmr hohmann2002
%
% References:
% C. F. Hauth, S. C. Berning, B. Kollmeier, and T. Brand. Modeling
% binaural unmasking of speech using a blind binaural processing stage.
% Trends in Hearing, 2020.
%
%
% Url: http://amtoolbox.org/amt-1.1.0/doc/models/hauth2020.php
% Copyright (C) 2009-2021 Piotr Majdak, Clara Hollomey, and the AMT team.
% This file is part of Auditory Modeling Toolbox (AMT) version 1.1.0
%
% 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/>.
% #StatusDoc: Good
% #StatusCode: Good
% #Verification: Unknown
% #Requirements: MATLAB M-Signal
% #Author: Christopher F. Hauth <christopher.hauth@uni-oldenburg.de>
% #Author: Dr. Thomas Brand <thomas.brand@uni-oldenburg.de>
% Date: 22.10.2020
%--------------------------------------------------------------------------
if nargin<2
error('%s: Too few input parameters.',upper(mfilename));
end
definput.import = {'hauth2020'}; % load defaults from arg_hauth2020
[flags,kv] = ltfatarghelper({},definput,varargin);
fs_model = 44100; % model sampling rate (must be 44100 Hz)
iNumofOptsigs = size(kv.OptSigs,2)/2;
frange = [kv.fmin kv.fmax];
BMtype = kv.Filterbank;
ERB_factor = kv.ERB_factor;
if flags.do_longterm
batch_flag = 1;
else
batch_flag = 0;
end
if flags.do_binauralinaccuracies
bin_inaccuracy = 1;
else
bin_inaccuracy = 0;
end
if frange(1)==frange(2)
ftarget = frange(1);
else
ftarget = kv.f_target; %[Hz]
end
% resmaple input if necessary
if fs_model~=fs
MixedSig = resample(MixedSig,fs_model,fs);
if ~isempty(kv.OptSigs)
kv.OptSigs = resample(kv.OptSigs,fs_model,fs);
end
end
% define temporal windows for EC processing and better ear selection
% batch processing means that the whole signal is considered as a single time frame
if batch_flag
iBlockLen_EC = length(MixedSig);
iBlockLen_BE = length(MixedSig);
iNumofBlocks_EC = 1;
iNumofBlocks_BE = 1;
else
% If short time processing is performed, EC process and better ear
% processing operate on different time constants.
% The EC process is conducted in
% time frames of 300ms (according to Hauth&Brand, 2018)
% Using a frame shift of 50% leads to an effective temporal window of
% 150 ms.
iBlockLen_EC = round(0.300.*fs_model);
% Better ear processing is performed on time frames of 20ms
% Using a frame shift of 50% leads to an effective temporal window of
% 10 ms.
iBlockLen_BE = round(0.020.*fs_model);
% make sure the block length is even
if mod(iBlockLen_EC,2)
iBlockLen_EC = iBlockLen_EC+1;
end
iNumofBlocks_EC = 2.*floor(length(MixedSig)./iBlockLen_EC)-1;
iNumofBlocks_BE = 2.*floor(length(MixedSig)./iBlockLen_BE)-1;
iBlockshift_EC = floor(iBlockLen_EC/2); % 50 percent overlap
iBlockshift_BE = floor(iBlockLen_BE/2);
window_EC = hann(iBlockLen_EC,'periodic');
window_BE = hann(iBlockLen_BE,'periodic');
end
% Binaural Processing Inaccuracies defined by vom H�vel (1984)
% For details see supplementary material of Hauth et al. (2020) (not yet published)
if bin_inaccuracy
% error in level
sigma_epsilon0 = 1.5; % [dB]
alpha0 = 15; % [dB]
p = 1.6; % no unit as it used as exponent
% error in time
sigmadelta0 = 65.*10^-6; % [sec]
Delta0 = 1.6.*10^-3;% [sec]
else
sigmadelta0 = 0;% [sec]
Delta0 = 1;
end
% Parameters for the gammatone filterbank
order = 4;
filter_per_ERB = 1;
%parameter defining the cut-off frequency of EC processing. Above this
%frequency, better-ear processing is assumed to play the dominant role.
f_cut = 1500; % 1500 Hz
% Gammatone Filterbank from described in Hohmann (2002)
if strcmp(kv.Filterbank,'GT')
analyzer1 = hohmann2002(fs_model,frange(1),ftarget,frange(2),filter_per_ERB,order,ERB_factor);
%% Apply the Gammatone filterbank
% LEFT EAR
[MixsigLbatch, ~] = hohmann2002_process(analyzer1, MixedSig(:,1));
fc = analyzer1.center_frequencies_hz;
MixsigLbatch = real(MixsigLbatch);
SomeImagSig = imag(MixsigLbatch);
% RIGHT EAR
[MixsigRbatch, ~] = hohmann2002_process(analyzer1, MixedSig(:,2));
MixsigRbatch = real(MixsigRbatch);
% If optional signals are present, process in the same way as mixed
% signals
if ~isempty(kv.OptSigs)
OptSigs = kv.OptSigs.';
if ~isempty(kv.OptSigs)
for mm=1:iNumofOptsigs
eval(sprintf('[OptSig%dLbatch, ~] = hohmann2002_process(analyzer1, OptSigs(2*(mm-1)+1,:)'');',mm));
eval(sprintf('OptSig%dLbatch = (real(OptSig%dLbatch));',mm,mm));
eval(sprintf('[OptSig%dRbatch, ~] = hohmann2002_process(analyzer1, OptSigs(2*mm,:)'');',mm));
eval(sprintf('OptSig%dRbatch = (real(OptSig%dRbatch));',mm,mm));
%eval(sprintf('[OptSig%dLbatchH, ~] = gfb_analyzer_process(analyzer1, OptSigs(:,2*(mm-1)+1));',mm));
%eval(sprintf('OptSig%dLbatch = transpose(real(OptSig%dLbatchH));',mm,mm));
%eval(sprintf('[OptSig%dRbatchH, ~] = gfb_analyzer_process(analyzer1, OptSigs(:,2*mm));',mm));
%eval(sprintf('OptSig%dRbatch = transpose(real(OptSig%dRbatchH));',mm,mm));
end
OptSigs = OptSigs.';
end
end
end
% Allocate buffer for the final resynthesis of signals
% the ending fin denotes final representation of the signal
MixsigLfin = zeros(size(MixsigLbatch));
MixsigRfin = MixsigLfin;
Mixsigminfin = MixsigLfin;
Mixsigmaxfin = MixsigLfin;
Mixsigsynfin = MixsigLfin;
% Allocate buffer for optional signals if present:
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dLfin = MixsigLfin;',ll));
eval(sprintf('OptSig%dRfin = MixsigLfin;',ll));
eval(sprintf('OptSig%dminfin = MixsigLfin;',ll));
eval(sprintf('OptSig%dmaxfin = MixsigLfin;',ll));
eval(sprintf('OptSig%dsynfin = MixsigLfin;',ll));
end
end
%% Binaural stage processing starts here:
% iterate through frequencies
for kk = 1:length(analyzer1.center_frequencies_hz)
% Check, if frequency is in the range of frequencies where binaural
% processing is assumed
if analyzer1.center_frequencies_hz(kk)<=f_cut
% Iterate through time blocks for EC processing
for mm = 1:iNumofBlocks_EC
if iNumofBlocks_EC==1 % i.e. considering the signal as a single time frame
% save signal portion
MixsigL = MixsigLbatch(:,kk);
MixsigR = MixsigRbatch(:,kk);
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dL = OptSig%dLbatch(:,kk);',ll,ll));
eval(sprintf('OptSig%dR = OptSig%dRbatch(:,kk);',ll,ll));
end
end
else
% Do block-wise processing
MixsigL = MixsigLbatch((mm-1)*iBlockshift_EC+1:...
iBlockLen_EC+(mm-1)*iBlockshift_EC,kk).*sqrt(window_EC);
MixsigR = MixsigRbatch((mm-1)*iBlockshift_EC+1:...
iBlockLen_EC+(mm-1)*iBlockshift_EC,kk).*sqrt(window_EC);
% Do the same for optional signals if present
%if exist('OptSigs','Var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dL = OptSig%dLbatch((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk).*sqrt(window_EC);',ll,ll));
eval(sprintf('OptSig%dR = OptSig%dRbatch((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk).*sqrt(window_EC);',ll,ll));
end
end
end
% Get levels of left and right ear signal
rmsL = sqrt(mean(MixsigL.^2));
rmsR = sqrt(mean(MixsigR.^2));
% calculate mean between ears. This will be the target level
% for the re-synthesis
rms_target = mean([rmsL' rmsR'],2);
% Level difference between left and right ear.
alpha = 20.*log10(rmsL./rmsR);
% Calculate level error (depending on the level difference)
if bin_inaccuracy
sigma_epsilon = sigma_epsilon0.*(1+(abs(alpha)/alpha0).^p);
levErrorL = mean(10.^(sigma_epsilon.*randn(1,1)./20));
levErrorR = mean(10.^(sigma_epsilon.*randn(1,1)./20));
else
% if processing is assumed to be perfect:
levErrorL = 1;
levErrorR = 1;
end
% Calculate factors for level equalization:
facR = sqrt((rmsL./rmsR).*levErrorR*levErrorL);
facL = 1./facR;
%% Apply equalization in Level (First step of EC Process)
MixsigLproc = facL.*MixsigL;
MixsigRproc = facR.*MixsigR;
% Apply the same equalization to the Optional Signals:
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dLproc=facL.*OptSig%dL;',ll,ll));
eval(sprintf('OptSig%dRproc=facR.*OptSig%dR;',ll,ll));
end
end
%% Calculate delays to maximize and minimize the output of the EC
% mechanism
% run EC processing for the mixed signals:
[Mixsigmin, Mixsigmax, ECparam4OptSigs] = hauth2020_fftcon(MixsigRproc,MixsigLproc,...
analyzer1.center_frequencies_hz(kk),fs_model,sigmadelta0,Delta0,bin_inaccuracy);
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('[OptSig%dmin,OptSig%dmax] = hauth2020_ecprocess4optsigs(OptSig%dRproc,OptSig%dLproc,ECparam4OptSigs,fs_model,bin_inaccuracy);',ll,ll,ll,ll));
end
end
rms_max = rms(Mixsigmax);
Mixsigmax = rms_target./rms_max.*Mixsigmax;
rms_min = rms(Mixsigmin);
Mixsigmin = rms_target./rms_min.*Mixsigmin;
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dmin = rms_target./rms_min.*OptSig%dmin;',ll,ll));
eval(sprintf('OptSig%dmax = rms_target./rms_max.*OptSig%dmax;',ll,ll));
end
end
% Use the modified SRMR and apply it to both EC processed signals
[ratiomax(mm,kk), ~] = hauth2020_srmr(Mixsigmax, fs_model, 'fast', 0, 'norm', 1, 'minCF', 4, 'maxCF', 32, 'single',0,'window',iBlockLen_EC);
[ratiomin(mm,kk), ~] = hauth2020_srmr(Mixsigmin, fs_model, 'fast', 0, 'norm', 1, 'minCF', 4, 'maxCF', 32, 'single',0,'window',iBlockLen_EC);
if mm>2
if mean([ratiomax(mm,kk) ratiomax(mm-1,kk)])>mean([ratiomin(mm,kk) ratiomin(mm-1,kk)]);
Mixsigsyn = Mixsigmax;
%if exist('OptSigs','Var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dsyn = OptSig%dmax;',ll,ll));
end
end
amt_disp(['Max used in band ',num2str(kk)],flags.disp);
else
Mixsigsyn = Mixsigmin;
%if exist('OptSigs','Var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dsyn = OptSig%dmin;',ll,ll));
end
end
amt_disp(['Min used in band ',num2str(kk)],flags.disp);
end
else
if ratiomax(mm,kk)>ratiomin(mm,kk)
Mixsigsyn = Mixsigmax;
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dsyn = OptSig%dmax;',ll,ll));
end
end
amt_disp(['Max used in band ',num2str(kk)],flags.disp);
else
Mixsigsyn = Mixsigmin;
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dsyn = OptSig%dmin;',ll,ll));
end
end
amt_disp(['Min used in band ',num2str(kk)],flags.disp);
end
end
% As the SII was used in the study of Hauth et al.(2020), the levels of the optional
% signals (if present) are saved
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('LevelOptSig%dmin(kk) = 20.*log10(rms(OptSig%dmin));',ll,ll));
eval(sprintf('LevelOptSig%dmax(kk) = 20.*log10(rms(OptSig%dmax));',ll,ll));
eval(sprintf('LevelOptSig%dsyn(kk) = 20.*log10(rms(OptSig%dsyn));',ll,ll));
eval(sprintf('LevelOptSig%dL(kk) = 20.*log10(rms(OptSig%dL));',ll,ll));
eval(sprintf('LevelOptSig%dR(kk) = 20.*log10(rms(OptSig%dR));',ll,ll));
end
end
% Construct final signals for resynthesis
if iNumofBlocks_EC==1
Mixsigminfin(:,kk) = Mixsigminfin(:,kk) + Mixsigmin;
Mixsigmaxfin(:,kk) = Mixsigmaxfin(:,kk) + Mixsigmax;
Mixsigsynfin(:,kk) = Mixsigsynfin(:,kk)+ Mixsigsyn;
MixsigLfin(:,kk) = MixsigLfin(:,kk) + (rms_target./rmsL).* MixsigL;
MixsigRfin(:,kk) = MixsigRfin(:,kk) + (rms_target./rmsR).*MixsigR;
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dminfin(:,kk) = OptSig%dminfin(:,kk) + OptSig%dmin;',ll,ll,ll));
eval(sprintf('OptSig%dmaxfin(:,kk) = OptSig%dmaxfin(:,kk) + OptSig%dmax;',ll,ll,ll));
eval(sprintf('OptSig%dsynfin(:,kk) = OptSig%dsynfin(:,kk) + OptSig%dsyn;',ll,ll,ll));
eval(sprintf('OptSig%dLfin(:,kk) = OptSig%dLfin(:,kk) +(rms_target./rmsL).* OptSig%dL;',ll,ll,ll));
eval(sprintf('OptSig%dRfin(:,kk) = OptSig%dRfin(:,kk) +(rms_target./rmsR).* OptSig%dR;',ll,ll,ll));
end
end
else
Mixsigminfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)...
= Mixsigminfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+...
Mixsigmin.*sqrt(window_EC);
Mixsigmaxfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)...
= Mixsigmaxfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+...
Mixsigmax.*sqrt(window_EC);
Mixsigsynfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)...
= Mixsigsynfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+...
Mixsigsyn.*sqrt(window_EC);
MixsigLfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)...
= MixsigLfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+...
MixsigL.*sqrt(window_EC);
MixsigRfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)...
= MixsigRfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+...
MixsigR.*sqrt(window_EC);
% Do the same for optional signals if present
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dminfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk) = OptSig%dminfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+OptSig%dmin.*sqrt(window_EC);',ll,ll,ll));
eval(sprintf('OptSig%dmaxfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk) = OptSig%dmaxfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+OptSig%dmax.*sqrt(window_EC);',ll,ll,ll));
eval(sprintf('OptSig%dsynfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk) = OptSig%dsynfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+OptSig%dsyn.*sqrt(window_EC);',ll,ll,ll));
eval(sprintf('OptSig%dLfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk) = OptSig%dLfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+OptSig%dL.*sqrt(window_EC);',ll,ll,ll));
eval(sprintf('OptSig%dRfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk) = OptSig%dRfin((mm-1)*iBlockshift_EC+1:iBlockLen_EC+(mm-1)*iBlockshift_EC,kk)+OptSig%dR.*sqrt(window_EC);',ll,ll,ll));
end
end
end
end
else
% Better ear processing:
for mm = 1:iNumofBlocks_BE
% Save left and right ear signal to temporary processing
% variables.
% if batch processing is used (whole signal is a single time
% frame), save the whole signal.
if iNumofBlocks_BE==1
MixsigL = MixsigLbatch(:,kk);
MixsigR = MixsigRbatch(:,kk);
% Do the same for the optional signals.
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dL = OptSig%dLbatch(:,kk);',ll,ll));
eval(sprintf('OptSig%dR = OptSig%dRbatch(:,kk);',ll,ll));
end
end
else
% if short time processing is used, save only a signal portion.
MixsigL = MixsigLbatch((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk).*sqrt(window_BE);
MixsigR = MixsigRbatch((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk).*sqrt(window_BE);
% Do the same for the optional signals.
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dL = OptSig%dLbatch((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk).*sqrt(window_BE);',ll,ll));
eval(sprintf('OptSig%dR = OptSig%dRbatch((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk).*sqrt(window_BE);',ll,ll));
end
end
end
rmsL = sqrt(mean(MixsigL.^2));
rmsR = sqrt(mean(MixsigR.^2));
rms_target = mean([rmsL' rmsR'],2);
% Use the modified SRMR and apply it to the left and right ear
% signal to identify the better ear.
[ratioL, ~] = hauth2020_srmr(rms_target./rmsL.*MixsigL, fs_model, 'fast', 0, 'norm', 1, 'minCF', 4, 'maxCF', 32, 'single',0,'window',iBlockLen_BE);
[ratioR, ~] = hauth2020_srmr(rms_target./rmsR.*MixsigR, fs_model, 'fast', 0, 'norm', 1, 'minCF', 4, 'maxCF', 32, 'single',0,'window',iBlockLen_BE);
% Independent of the EC processing strategy, the better ear is
% always selected, and the worse ear is always neglected.
% Therefore, the better ear channels are combined with all EC
% outputs. (Min+BE, Max+BE, Syn+BE).
if ratioL>=ratioR
% if the left ear channel has stronger modulation cues,
% select it set it to the target level.
Mixsigmin = rms_target./rmsL.*MixsigL;
Mixsigmax = Mixsigmin;
Mixsigsyn = Mixsigmin;
% Do the same for optional signals
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dsyn = rms_target./rmsL.*OptSig%dL;',ll,ll));
eval(sprintf('OptSig%dmin = rms_target./rmsL.*OptSig%dL;',ll,ll));
eval(sprintf('OptSig%dmax = rms_target./rmsL.*OptSig%dL;',ll,ll));
end
end
amt_disp('Left Ear used', flags.disp);
else
% if the right ear channel has stronger modulation cues,
% select it and set it to the target level.
Mixsigmin = rms_target./rmsR.*MixsigR;
Mixsigmax = Mixsigmin;
Mixsigsyn = Mixsigmin;
% Do the same for optional signals
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dsyn = rms_target./rmsR.*OptSig%dR;',ll,ll));
eval(sprintf('OptSig%dmin = rms_target./rmsR.*OptSig%dR;',ll,ll));
eval(sprintf('OptSig%dmax = rms_target./rmsR.*OptSig%dR;',ll,ll));
end
end
amt_disp('Right Ear used', flags.disp);
end
% Because the SII was used in the study by Hauth et al.(2020),
% the frequency specific levels of the optional signals are
% calculated. They can later be used by the SII.
% If no optional signals are present, the levels are not
% calculated.
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('LevelOptSig%dmin(kk) = 20.*log10(rms(OptSig%dmin));',ll,ll));
eval(sprintf('LevelOptSig%dmax(kk) = 20.*log10(rms(OptSig%dmax));',ll,ll));
eval(sprintf('LevelOptSig%dsyn(kk) = 20.*log10(rms(OptSig%dsyn));',ll,ll));
eval(sprintf('LevelOptSig%dL(kk) = 20.*log10(rms(OptSig%dL));',ll,ll));
eval(sprintf('LevelOptSig%dR(kk) = 20.*log10(rms(OptSig%dR));',ll,ll));
end
end
% save the channels in the final output matrices.
if iNumofBlocks_BE==1
% Minimized output
Mixsigminfin(:,kk)= Mixsigminfin(:,kk) + Mixsigmin;
% Maximized output
Mixsigmaxfin(:,kk) = Mixsigmaxfin(:,kk)+ Mixsigmax;
% Synthesized output
Mixsigsynfin(:,kk) = Mixsigsynfin(:,kk) + Mixsigsyn;
% Left ear output
MixsigLfin(:,kk) = MixsigLfin(:,kk) +(rms_target./rmsL).*MixsigL;
% Right ear output
MixsigRfin(:,kk) = MixsigRfin(:,kk) +(rms_target./rmsR).*MixsigR;
% Do the same for optional signals.
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dminfin(:,kk) = OptSig%dminfin(:,kk) + OptSig%dmin;',ll,ll,ll));
eval(sprintf('OptSig%dmaxfin(:,kk) = OptSig%dmaxfin(:,kk) + OptSig%dmax;',ll,ll,ll));
eval(sprintf('OptSig%dsynfin(:,kk) = OptSig%dsynfin(:,kk) + OptSig%dsyn;',ll,ll,ll));
eval(sprintf('OptSig%dLfin(:,kk) = OptSig%dLfin(:,kk) +(rms_target./rmsL).* OptSig%dL;',ll,ll,ll));
eval(sprintf('OptSig%dRfin(:,kk) = OptSig%dRfin(:,kk) +(rms_target./rmsR).* OptSig%dR;',ll,ll,ll));
end
end
% save channels to final output matrices ( if short time processing was used)
else
%Mixed Signals
% Mimizized
Mixsigminfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)...
= Mixsigminfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)+Mixsigmin.*sqrt(window_BE);
% Maximimized
Mixsigmaxfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)...
= Mixsigmaxfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)+ Mixsigmax.*sqrt(window_BE);
% Synthesized
Mixsigsynfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)...
= Mixsigsynfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)+ Mixsigsyn.*sqrt(window_BE);
%Left
MixsigLfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)...
= MixsigLfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)+MixsigL.*sqrt(window_BE);
%Right
MixsigRfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)...
= MixsigRfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+...
(mm-1)*iBlockshift_BE,kk)+MixsigR.*sqrt(window_BE);
% Process Optional Signals accordingly:
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dminfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk) = OptSig%dminfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk)+OptSig%dmin.*sqrt(window_BE);',ll,ll,ll));
eval(sprintf('OptSig%dmaxfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk) = OptSig%dmaxfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk)+OptSig%dmax.*sqrt(window_BE);',ll,ll,ll));
eval(sprintf('OptSig%dsynfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk) = OptSig%dsynfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk)+OptSig%dsyn.*sqrt(window_BE);',ll,ll,ll));
eval(sprintf('OptSig%dLfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk) = OptSig%dLfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk)+OptSig%dL.*sqrt(window_BE);',ll,ll,ll));
eval(sprintf('OptSig%dRfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk) = OptSig%dRfin((mm-1)*iBlockshift_BE+1:iBlockLen_BE+(mm-1)*iBlockshift_BE,kk)+OptSig%dR.*sqrt(window_BE);',ll,ll,ll));
end
end
end
end
end
end
%% Resynthesize time domain signals using gammatone filters and a delay of 10ms
synthesizer = hohmann2002_synth(analyzer1,0.01);%1/fs_model);%1/fs_model);
Mixsigminfin = hohmann2002_process(synthesizer, Mixsigminfin+SomeImagSig)';
Mixsigmaxfin= hohmann2002_process(synthesizer, Mixsigmaxfin)';
Mixsigsynfin = hohmann2002_process(synthesizer, Mixsigsynfin)';
MixsigLfin = hohmann2002_process(synthesizer,MixsigLfin+SomeImagSig)';
MixsigRfin = hohmann2002_process(synthesizer,MixsigRfin+SomeImagSig)';
% Do the same for optional signals
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('OptSig%dminfin = hohmann2002_process(synthesizer, (OptSig%dminfin)+SomeImagSig)'';',ll,ll));
eval(sprintf('OptSig%dmaxfin = hohmann2002_process(synthesizer, (OptSig%dmaxfin))'';',ll,ll));
eval(sprintf('OptSig%dsynfin = hohmann2002_process(synthesizer, (OptSig%dsynfin))'';',ll,ll));
eval(sprintf('OptSig%dLfin = hohmann2002_process(synthesizer, (OptSig%dLfin)+SomeImagSig)'';',ll,ll));
eval(sprintf('OptSig%dRfin = hohmann2002_process(synthesizer, (OptSig%dRfin)+SomeImagSig)'';',ll,ll));
end
end
%% Resample signals if necessary
if fs_model~= fs
Mixsigminfin = resample(Mixsigminfin,fs,fs_model);
Mixsigmaxfin = resample(Mixsigmaxfin,fs,fs_model);
Mixsigsynfin = resample(Mixsigsynfin,fs,fs_model);
MixsigLfin = resample(MixsigLfin,fs,fs_model);
MixsigRfin = resample(MixsigRfin,fs,fs_model);
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('[OptSig%dminfin, ~] = resample(OptSig%dminfin,fs,fs_model);',ll,ll));
eval(sprintf('[OptSig%dmaxfin, ~] = resample(OptSig%dmaxfin,fs,fs_model);',ll,ll));
eval(sprintf('[OptSig%dsynfin, ~] = resample(OptSig%dsynfin,fs,fs_model);',ll,ll));
eval(sprintf('[OptSig%dLfin, ~] = resample(OptSig%dLfin,fs,fs_model);',ll,ll));
eval(sprintf('[OptSig%dRfin, ~] = resample(OptSig%dRfinfs,fs_model);',ll,ll));
end
end
end
%% Save signals to output structure
out_struct.signals.MixsigSyn = Mixsigsynfin;
out_struct.signals.MixsigMin = Mixsigminfin;
out_struct.signals.MixsigMax = Mixsigmaxfin;
out_struct.signals.MixsigL = MixsigLfin;
out_struct.signals.MixsigR = MixsigRfin;
% Save also the optional signals to the output structure
%if exist('OptSigs','var')
if ~isempty(kv.OptSigs)
for ll=1:iNumofOptsigs
eval(sprintf('out_struct.signals.OptSig%dsyn = OptSig%dsynfin;',ll,ll));
eval(sprintf('out_struct.levels.LevelOptSig%dsyn = LevelOptSig%dsyn;',ll,ll));
eval(sprintf('out_struct.signals.OptSig%dmin = OptSig%dminfin;',ll,ll));
eval(sprintf('out_struct.levels.LevelOptSig%dmin = LevelOptSig%dmin;',ll,ll));
eval(sprintf('out_struct.signals.OptSig%dmax = OptSig%dmaxfin;',ll,ll));
eval(sprintf('out_struct.levels.LevelOptSig%dmax = LevelOptSig%dmax;',ll,ll));
eval(sprintf('out_struct.signals.OptSig%dL = OptSig%dLfin;',ll,ll));
eval(sprintf('out_struct.levels.LevelOptSig%dL = LevelOptSig%dL;',ll,ll));
eval(sprintf('out_struct.signals.OptSig%dR = OptSig%dRfin;',ll,ll));
eval(sprintf('out_struct.levels.LevelOptSig%dR = LevelOptSig%dR;',ll,ll));
end
end
%--------------------Licence ---------------------------------------------
% Copyright (c) <2020> Christopher F. Hauth
% Dept. Medical Physics and Acoustics
% Carl von Ossietzky University Oldenburg
% Permission is hereby granted, free of charge, to any person obtaining
% a copy of this software and associated documentation files
% (the "Software"), to deal in the Software without restriction, including
% without limitation the rights to use, copy, modify, merge, publish,
% distribute, sublicense, and/or sell copies of the Software, and to
% permit persons to whom the Software is furnished to do so, subject
% to the following conditions:
% The above copyright notice and this permission notice shall be included
% in all copies or substantial portions of the Software.
% THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
% EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
% OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
% IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
% CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
% TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
% SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
% END OF FILE
Build with Bootstrap