function varargout=audspecgram(insig,fs,varargin)
%AUDSPECGRAM Auditory spectrogram
% Usage: audspecgram(insig,fs,op1,op2, ... );
% C=audspecgram(insig,fs, ... );
%
% AUDSPECGRAM(insig,fs) plots an auditory spectrogram of the signal insig,
% which has been sampled at a sampling rate of fs Hz. The output is
% low-pass modulation filtered before presentation.
%
% The frequency axis is diplayed on a erb-scale, but labelled in
% Hz. Using the mouse to get plot coordinates will reveal the real
% value in erb's. Use erbtofreq to convert to Hz.
%
% C=AUDSPECGRAM(insig,fs, ... ) returns the image to be displayed as a
% matrix. Use this in conjunction with imwrite etc. Do *not* use this as a
% method to compute an auditory representation. Use some of the model
% preprocessing functions for this.
%
% Additional arguments can be supplied like this:
%
% audspecgram(insig,fs,'dynrange',30);
%
% The arguments must be character strings possibly followed by an argument:
%
% 'adapt' Model adaptation. This is the default. This options also
% sets the output to be displayed on a linear scale.
%
% 'noadapt' Do not model adaptation. This option also sets a dB scale to
% display the output.
%
% 'ihc',modelname
% Pass modelname to IHCENVELOPE to determine the inner
% hair cell envelope extraction process to use. Default is to
% use the 'dau' model.
%
% 'classic' Display a classic spectrogram. This option is equal to
% {'ihc','hilbert', 'noadapt', 'nomf'}
%
% 'mlp',f Modulation low-pass filter to frequency f. Default is to
% low-pass filter to 50 Hz.
%
% 'mf',f Modulation filter with specified center frequency.
%
% 'nomf' No modulation filtering of any kind.
%
% 'image' Use imagesc to display the spectrogram. This is the default.
%
% 'clim',clim Use a colormap ranging from clim(1) to clim(2). These
% values are passed to imagesc. See the help on imagesc.
%
% 'dynrange',r Limit the displayed dynamic range to r. This option
% is especially usefull when displaying on a dB scale (no adaptation).
%
% 'fullrange' Use the full dynamic range. This is the default.
%
% 'ytick' A vector containing the frequency in Hz of the yticks.
%
% 'thr',r Keep only the largest fraction r of the coefficients, and
% set the rest to zero.
%
% 'frange',frange
% Choose a frequency scale ranging from frange(1) to
% frange(2), values are entered in Hz. Default is to display from
% 0 to 8000 Hz.
%
% 'xres',xres Approximate number of pixels along x-axis / time.
%
% 'yres',yres Approximate number of pixels along y-axis / frequency If
% only one of 'xres' and 'yres' is specified, the default
% aspect ratio will be used.
%
% 'displayratio',r Set the default aspect ratio.
%
% 'contour' Do a contour plot to display the spectrogram.
%
% 'surf' Do a surf plot to display the spectrogram.
%
% 'mesh' Do a mesh plot to display the spectrogram.
%
% 'colorbar' Display the colorbar. This is the default.
%
% 'nocolorbar' Do not display the colorbar.
%
% Examples:
% ---------
%
% The following figure shows a classic spectrogram on an Erb scale of a
% spoken word:
%
% audspecgram(greasy,16000,'classic','dynrange',50);
%
% The next example shows a Dau-style spectrogram (including adaptation
% and modulation low-pass filtering) of the same word:
%
% audspecgram(greasy,16000);
%
% See also: dau1997_preproc
%
% Url: http://amtoolbox.sourceforge.net/amt-0.9.8/doc/general/audspecgram.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 : Peter Søndergaard.
if nargin<2
error('Too few input arguments.');
end;
if ~isnumeric(insig) || ~isvector(insig)
error('%s: Input must be a vector.',upper(mfilename));
end;
definput.import={'plotfilterbank','ltfattranslate','tfplot'};
definput.importdefaults={'lin','audtick'};
% Define initial value for flags and key/value pairs.
definput.flags.adapt={'adapt','noadapt'};
definput.flags.thr={'nothr','thr'};
definput.flags.mlp={'mlp','mf','nomf'};
definput.flags.delay={'gammatonedelay','zerodelay'};
definput.keyvals.ihc='ihc_dau';
definput.keyvals.thr=0;
%definput.keyvals.ytick=[0,100,250,500,1000,2000,4000,8000];
definput.keyvals.mlp=50;
definput.keyvals.mf=[0 5 10 16.6 27.7];
definput.keyvals.frange=[0,8000];
definput.keyvals.xres=800;
definput.keyvals.yres=600;
definput.flags.colorbar={'colorbar','nocolorbar'};
definput.groups.classic={'ihc','hilbert', 'noadapt', 'nomf'};
[flags,kv]=ltfatarghelper({},definput,varargin);
siglen=length(insig);
fhigh=kv.frange(2);
flow =kv.frange(1);
audlimits=freqtoerb(kv.frange);
% fhigh can at most be the Nyquest frequency
fhigh=min(fhigh,fs/2);
% Downsample this signal if it is sampled at a much higher rate than
% 2*fhigh. This reduces memory consumption etc. 1.5 and 1.2 are choosen as a
% safeguard to not loose information.
if fs>2*1.5*fhigh
fsnew=round(fhigh*2*1.2);
% Determine new signal length
siglen=round(siglen/fs*fsnew);
% Do the resampling using an FFT based method, as this is more flexible
% than the 'resample' method included in Matlab
insig=fftresample(insig,siglen);
% Switch to new value
fs=fsnew;
end;
% Determine the hopsize
% Using a hopsize different from 1 is currently not possible because all
% the subsequent filters fail because of a to low subband sampling rate.
%hopsize=max(1,floor(siglen/xres));
hopsize=1;
% find the center frequencies used in the filterbank
fc = erbspace(flow,fhigh,kv.yres);
if 1
% Calculate filter coefficients for the gammatone filter bank.
[gt_b, gt_a, delay]=gammatone(fc, fs, 'complex');
% Apply the Gammatone filterbank
outsig = 2*real(ufilterbankz(gt_b,gt_a,insig,hopsize));
else
L=siglen;
bw_gauss=audfiltbw(fc)/fs*L/0.79;
fc_gauss=round(fc/fs*L);
g=cell(1,kv.yres);
for m=1:kv.yres
g{m}=real(pgauss(L,'bw',bw_gauss(m),'cf',fc_gauss(m)));
end;
outsig=filterbank(insig,g,hopsize);
end;
% The subband are now (possibly) sampled at a lower frequency than the
% original signal.
fssubband=round(fs/hopsize);
outsig = ihcenvelope(outsig,fssubband,kv.ihc);
if flags.do_adapt
% non-linear adaptation loops
outsig = adaptloop(outsig, fssubband);
end;
if flags.do_nomf
modfilt_outsig=outsig;
end;
if flags.do_mlp
% Calculate filter coefficients for the 50 Hz modulation lowpass
% filter. Just use a 2nd order Butterworth for this.
% FIXME: This filter places a pole /very/ close to the unit circle.
%mlp_a = exp(-(1/0.02)/fs);
mlp_a = exp(-kv.mlp/fs);
mlp_b = 1 - mlp_a;
mlp_a = [1, -mlp_a];
% Apply the low-pass modulation filter.
modfilt_outsig = filter(mlp_b,mlp_a,outsig);
end;
if flags.do_mf
nreps=length(kv.mf)-1;
else
nreps=1;
end;
% Loop over the number of modulation frequency channels
for jj=1:nreps
if flags.do_mf
% Calculate filter coefficients for the 50 Hz modulation lowpass
% filter. Just use a 2nd order Butterworth for this.
[mf_b,mf_a] = butter(2,[kv.mf(jj),kv.mf(jj+1)]/(subbandfs/2));
% Apply the modulation filter.
modfilt_outsig = filter(mf_b,mf_a,outsig);
if mfc(nmfc) <= 10
modfilt_outsig = 1*real(modfilt_outsig);
else
modfilt_outsig = 1/sqrt(2)*abs(modfilt_outsig);
end
end;
if flags.do_thr
% keep only the largest coefficients.
modfilt_outsig=largestr(modfilt_outsig,kv.thr);
end
% Apply transformation to coefficients.
if flags.do_noadapt
% This is a safety measure to avoid log of negative numbers.
modfilt_outsig(:)=max(modfilt_outsig(:),eps);
modfilt_outsig=20*log10(modfilt_outsig);
end;
plotfilterbank(modfilt_outsig,1,fc,fs,'argimport',flags,kv,'fc',fc,'fs',fs);
if nargout>0
varargout={modfilt_outsig,fc};
end;
end
% complex frequency shifted first order lowpass
function [b,a] = efilt(w0,bw);
e0 = exp(-bw/2);
b = 1 - e0;
a = [1, -e0*exp(1i*w0)];