%EXAMP_SPEECH  Spectrogram of speech signal.%%   This example show how to work with almost real world data.%%   FIGURE 1 Linus%%    This show a spectrogram of Linus Thorvalds saying "My name is%    Linus Thorvalds, and I pronounce Linux as Linux"%%    The 90% lowest values have been filtered away, to produce a%    cleaner plot.%disp('Type "help examp_speech" to see a description of how this example works.');% Load the signalf=linus();% The parameters below are choosen such that the last part of the signal is% cut of; this does not matter, it is zero anyway. L=40320;a=224;M=280;b=L/M;% Optimally centered window.g=pgauss(L,a/b);% Get some coefficients.c=dgt(f,g,a,M,L);% Since the signal is real, then the absolute value of the% upper half of crect is a mirror image of the lower half. Therefore, we skip% it. cplot=abs(c(1:floor(M/2)+1,:)).^2;% Plot.figure(1);% Plotting is diffent in Matlab and Octave.if isoctave  title('Linus speaking.');  imagesc(flipud(largestr(cplot,.1)));else  contour(log(largestr(cplot,.1)+eps));  title('Linus speaking.');end;% XXL: Wouldn't it be better to have 'real' time and frequency scales?% use sgram here????% We need to figure out what this example should do.% XXL: For speech I was expecting sound output%if ~isoctave % maybe only working for matlab?%    if ispc%        wavplay(f,8000) % SR= 8000 estimated%    else % unix : does it work?%       % there should not be other systems, no?%        auwrite(f,8000,16,'/tmp/tmpsnd.au');%        unix('cat /tmp/tmpsnd.au >/dev/audio'); %        % % also possible:%        % wavwrite(SIG.sig/m,SIG.fs,'/tmp/tmp.wav');%        % unix('play /tmp/tmp.wav');%        % unix('rm -f /tmp/tmp.wav');   %    end%end