test_subfingerprint.m

fingerprinting test,it is very helpful to the learner

% zorro write test_subfingerprint
% Author: zorro
% Date: 2007-3-4
function [s1_subfingerprint,aspc] = test_subfingerprint(wavefile,method,startpos,length)

if nargin < 2;   method = 0; end
if nargin < 3;   startpos = 1;    end
if nargin < 4;   length = 90*1024; end 

% Load a speech waveform
 [d,sr] = wavread(wavefile, [startpos, startpos + length]); %'buddy.wav'); 
 
if 1 == method 
    d = decimate(d,2); sr = sr/2;  %decimate2
end
if 2 == method
    d = decimate(d,4); sr = sr/4;  %decimate4
end
if 3 == method
    d = decimate(d,8); sr = sr/8;  %decimate8
end
if 4 == method
    d = downsample(d,2); sr = sr/2;  %downsample2
end
if 5 == method
    d = downsample(d,4); sr = sr/4; %downsample4
end
if 6 == method
    N = randn(1,length(d));%guassis noise;
    d=d+N';
end
if 7 == method
    N=rand(1,length(d));%white noise;
    d=d+N';    
end
    
 subplot(521)
 
 plot(d);
 ylabel('Amplitude')
 title('original wave')

 % Look at its regular spectrogram
 subplot(523)
 
 specgram(d, 256, sr);
 title('original spectrogram')

 % Calculate basic RASTA-PLP cepstra and spectra
 [cep1, spec1] = rastaplp(d, sr);
 
 % Convert to MFCCs very close to those genrated by feacalc -sr 22050 -nyq
 % 8000 -dith -hpf -opf htk -delta 0 -plp no -dom cep -com yes -frq mel -filt tri -win 32 -step 16 -cep 20
 % [mm,aspc,pspc] = melfcc(d*3.3752, sr, 'minfreq', 300, 'maxfreq', 2000, 'numcep', 33, 'nbands', 33, 'fbtype', 'fcmel', 'dcttype', 1, 'usecmp', 1, 'wintime', 0.037, 'hoptime', 0.0116, 'preemph', 0, 'dither', 1);
[mm,aspc,pspc] = melfcc(d*3.3752, sr, 'minfreq', 300, 'maxfreq', 2000, 'sumpower',0, 'numcep', 33, 'nbands', 33, 'fbtype', 'bark', 'dcttype', 1, 'usecmp', 1, 'wintime', 0.037, 'hoptime', 0.0116, 'preemph', 0, 'dither', 1);

 % .. and plot them
 subplot(525)
 imagesc(10*log10(pspc)); % Power spectrum, so dB is 10log10
 title('Power spectrum features')
 ylabel('dB')
 axis xy
 
 subplot(527)
 imagesc(aspc)
  title('33-band bark filter-bank')
   xlabel('time/10ms frames')
 axis xy
 % Notice the auditory warping of the frequency axis to give more 
 % space to low frequencies and the way that RASTA filtering 
 % emphasizes the onsets of static sounds like vowels
 
 aspc=filter([1,-1],1,aspc);

 [dim,len] = size(aspc);

 s1_fingerprint = zeros(dim,32);

 s1_subfingerprint = zeros(256,32);
for n=1:len
for m=1:32
    if n==1
        t=aspc(m,n)-aspc(m+1,n);
    else
        t=aspc(m,n)-aspc(m+1,n)-aspc(m,n-1)-aspc(m+1,n-1);
    end
    if t>0
        if n<256
            s1_subfingerprint(n,m)=1;
        end
        s1_fingerprint(n,m)=1;
    end
end
end


 % .. and plot them
 subplot(529)
 imagesc(s1_subfingerprint);
 title('the first fingerprint band')

 axis xy

 subplot(5,2,[2 4 6 8 10])
 
 imagesc(s1_fingerprint);
 title('fingerprint')

  axis xy
  
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

 
end