|
boblsturm@0
|
1 function [p,m,S] = ifptrack(d,w,sr,fminl,fminu,fmaxl,fmaxu)
|
|
boblsturm@0
|
2 % [p,m,S] = ifptrack(d,w,sr,fminl,fminu,fmaxl,fmaxu)
|
|
boblsturm@0
|
3 % Pitch track based on inst freq.
|
|
boblsturm@0
|
4 % Look for adjacent bins with same inst freq.
|
|
boblsturm@0
|
5 % d is the input waveform. sr is its sample rate
|
|
boblsturm@0
|
6 % w is the basic STFT DFT length (window is half, hop is 1/4)
|
|
boblsturm@0
|
7 % S returns the underlying complex STFT.
|
|
boblsturm@0
|
8 % fmin,fmax define ramps at edge of sensitivity
|
|
boblsturm@0
|
9 % 2006-05-03 dpwe@ee.columbia.edu
|
|
boblsturm@0
|
10
|
|
boblsturm@0
|
11 % Copyright (c) 2006 Columbia University.
|
|
boblsturm@0
|
12 %
|
|
boblsturm@0
|
13 % This file is part of LabROSA-coversongID
|
|
boblsturm@0
|
14 %
|
|
boblsturm@0
|
15 % LabROSA-coversongID is free software; you can redistribute it and/or modify
|
|
boblsturm@0
|
16 % it under the terms of the GNU General Public License version 2 as
|
|
boblsturm@0
|
17 % published by the Free Software Foundation.
|
|
boblsturm@0
|
18 %
|
|
boblsturm@0
|
19 % LabROSA-coversongID is distributed in the hope that it will be useful, but
|
|
boblsturm@0
|
20 % WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
boblsturm@0
|
21 % MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
boblsturm@0
|
22 % General Public License for more details.
|
|
boblsturm@0
|
23 %
|
|
boblsturm@0
|
24 % You should have received a copy of the GNU General Public License
|
|
boblsturm@0
|
25 % along with LabROSA-coversongID; if not, write to the Free Software
|
|
boblsturm@0
|
26 % Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
|
|
boblsturm@0
|
27 % 02110-1301 USA
|
|
boblsturm@0
|
28 %
|
|
boblsturm@0
|
29 % See the file "COPYING" for the text of the license.
|
|
boblsturm@0
|
30
|
|
boblsturm@0
|
31 % downweight fundamentals below here
|
|
boblsturm@0
|
32 if nargin < 4; fminl = 150; end
|
|
boblsturm@0
|
33 if nargin < 5; fminu = 300; end
|
|
boblsturm@0
|
34 % highest frequency we look to
|
|
boblsturm@0
|
35 if nargin < 6; fmaxl = 2000; end
|
|
boblsturm@0
|
36 if nargin < 7; fmaxu = 4000; end
|
|
boblsturm@0
|
37
|
|
boblsturm@0
|
38
|
|
boblsturm@0
|
39 % Calculate the inst freq gram
|
|
boblsturm@0
|
40 [I,S] = ifgram(d,w,w/2,w/4,sr);
|
|
boblsturm@0
|
41
|
|
boblsturm@0
|
42 % Only look at bins up to 2 kHz
|
|
boblsturm@0
|
43 maxbin = round(fmaxu * (w/sr) );
|
|
boblsturm@0
|
44 %maxbin = size(I,1)
|
|
boblsturm@0
|
45 minbin = round(fminl * (w/sr) );
|
|
boblsturm@0
|
46
|
|
boblsturm@0
|
47 % Find plateaus in ifgram - stretches where delta IF is < thr
|
|
boblsturm@0
|
48 ddif = [I(2:maxbin, :);I(maxbin,:)] - [I(1,:);I(1:(maxbin-1),:)];
|
|
boblsturm@0
|
49
|
|
boblsturm@0
|
50 % expected increment per bin = sr/w, threshold at 3/4 that
|
|
boblsturm@0
|
51 dgood = abs(ddif) < .75*sr/w;
|
|
boblsturm@0
|
52
|
|
boblsturm@0
|
53 % delete any single bins (both above and below are zero);
|
|
boblsturm@0
|
54 dgood = dgood .* ([dgood(2:maxbin,:);dgood(maxbin,:)] > 0 | [dgood(1,:);dgood(1:(maxbin-1),:)] > 0);
|
|
boblsturm@0
|
55
|
|
boblsturm@0
|
56 % check it out
|
|
boblsturm@0
|
57 %p = dgood;
|
|
boblsturm@0
|
58
|
|
boblsturm@0
|
59 % reconstruct just pitchy cells?
|
|
boblsturm@0
|
60 %r = istft(p.*S,w,w/2,w/4);
|
|
boblsturm@0
|
61
|
|
boblsturm@0
|
62 p = 0*dgood;
|
|
boblsturm@0
|
63 m = 0*dgood;
|
|
boblsturm@0
|
64
|
|
boblsturm@0
|
65 % For each frame, extract all harmonic freqs & magnitudes
|
|
boblsturm@0
|
66 for t = 1:size(I,2)
|
|
boblsturm@0
|
67 ds = dgood(:,t)';
|
|
boblsturm@0
|
68 lds = length(ds);
|
|
boblsturm@0
|
69 % find nonzero regions in this vector
|
|
boblsturm@0
|
70 st = find(([0,ds(1:(lds-1))]==0) & (ds > 0));
|
|
boblsturm@0
|
71 en = find((ds > 0) & ([ds(2:lds),0] == 0));
|
|
boblsturm@0
|
72 npks = length(st);
|
|
boblsturm@0
|
73 frqs = zeros(1,npks);
|
|
boblsturm@0
|
74 mags = zeros(1,npks);
|
|
boblsturm@0
|
75 for i = 1:length(st)
|
|
boblsturm@0
|
76 bump = abs(S(st(i):en(i),t));
|
|
boblsturm@0
|
77 frqs(i) = (bump'*I(st(i):en(i),t))/(sum(bump)+(sum(bump)==0));
|
|
boblsturm@0
|
78 mags(i) = sum(bump);
|
|
boblsturm@0
|
79 if frqs(i) > fmaxu
|
|
boblsturm@0
|
80 mags(i) = 0;
|
|
boblsturm@0
|
81 frqs(i) = 0;
|
|
boblsturm@0
|
82 elseif frqs(i) > fmaxl
|
|
boblsturm@0
|
83 mags(i) = mags(i) * max(0, (fmaxu - frqs(i))/(fmaxu-fmaxl));
|
|
boblsturm@0
|
84 end
|
|
boblsturm@0
|
85 % downweight magnitudes below? 200 Hz
|
|
boblsturm@0
|
86 if frqs(i) < fminl
|
|
boblsturm@0
|
87 mags(i) = 0;
|
|
boblsturm@0
|
88 frqs(i) = 0;
|
|
boblsturm@0
|
89 elseif frqs(i) < fminu
|
|
boblsturm@0
|
90 % 1 octave fade-out
|
|
boblsturm@0
|
91 mags(i) = mags(i) * (frqs(i) - fminl)/(fminu-fminl);
|
|
boblsturm@0
|
92 end
|
|
boblsturm@0
|
93 if frqs(i) < 0
|
|
boblsturm@0
|
94 mags(i) = 0;
|
|
boblsturm@0
|
95 frqs(i) = 0;
|
|
boblsturm@0
|
96 end
|
|
boblsturm@0
|
97
|
|
boblsturm@0
|
98 end
|
|
boblsturm@0
|
99
|
|
boblsturm@0
|
100 % then just keep the largest at each frame (for now)
|
|
boblsturm@0
|
101 % [v,ix] = max(mags);
|
|
boblsturm@0
|
102 % p(t) = frqs(ix);
|
|
boblsturm@0
|
103 % m(t) = mags(ix);
|
|
boblsturm@0
|
104 % No, keep them all
|
|
boblsturm@0
|
105 %bin = st;
|
|
boblsturm@0
|
106 bin = round((st+en)/2);
|
|
boblsturm@0
|
107 p(bin,t) = frqs;
|
|
boblsturm@0
|
108 m(bin,t) = mags;
|
|
boblsturm@0
|
109 end
|
|
boblsturm@0
|
110
|
|
boblsturm@0
|
111 %% Pull out the max in each column
|
|
boblsturm@0
|
112 %[mm,ix] = max(m);
|
|
boblsturm@0
|
113 %% idiom to retrieve different element from each column
|
|
boblsturm@0
|
114 %[nr,nc] = size(p);
|
|
boblsturm@0
|
115 %pp = p((nr*[0:(nc-1)])+ix);
|
|
boblsturm@0
|
116 %mm = m((nr*[0:(nc-1)])+ix);
|
|
boblsturm@0
|
117 % r = synthtrax(pp,mm,sr,w/4);
|
|
boblsturm@0
|
118
|
|
boblsturm@0
|
119 %p = pp;
|
|
boblsturm@0
|
120 %m = mm;
|
|
boblsturm@0
|
121
|
