Mercurial > hg > camir-aes2014

annotate toolboxes/MIRtoolbox1.3.2/MIRToolbox/@mirchromagram/mirchromagram.m @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
rev   line source
wolffd@0 1 function varargout = mirchromagram(orig,varargin)
wolffd@0 2 % c = mirchromagram(x) computes the chromagram, or distribution of energy
wolffd@0 3 % along pitches, of the audio signal x.
wolffd@0 4 % (x can be the name of an audio file as well, or a spectrum, ...)
wolffd@0 5 % Optional argument:
wolffd@0 6 % c = mirchromagram(...,'Tuning',t): specifies the central frequency
wolffd@0 7 % (in Hz.) associated to chroma C.
wolffd@0 8 % Default value, t = 261.6256 Hz
wolffd@0 9 % c = mirchromagram(...,'Wrap',w): specifies whether the chromagram is
wolffd@0 10 % wrapped or not.
wolffd@0 11 % w = 1: groups all the pitches belonging to same pitch classes
wolffd@0 12 % (default value)
wolffd@0 13 % w = 0: pitches are considered as absolute values.
wolffd@0 14 % c = mirchromagram(...,'Frame',l,h) orders a frame decomposition of window
wolffd@0 15 % length l (in seconds) and hop factor h, expressed relatively to
wolffd@0 16 % the window length. For instance h = 1 indicates no overlap.
wolffd@0 17 % Default values: l = .2 seconds and h = .05
wolffd@0 18 % c = mirchromagram(...,'Center'): centers the result.
wolffd@0 19 % c = mirchromagram(...,'Normal',n): performs a n-norm of the
wolffd@0 20 % resulting chromagram. Toggled off if n = 0
wolffd@0 21 % Default value: n = Inf (corresponding to a normalization by
wolffd@0 22 % the maximum value).
wolffd@0 23 % c = mirchromagram(...,'Pitch',p): specifies how to label chromas in
wolffd@0 24 % the figures.
wolffd@0 25 % p = 1: chromas are labeled using pitch names (default)
wolffd@0 26 % alternative syntax: chromagram(...,'Pitch')
wolffd@0 27 % p = 0: chromas are labeled using MIDI pitch numbers
wolffd@0 28 % c = mirchromagram(...,'Triangle'): weight the contribution of each
wolffd@0 29 % frequency with respect to the distance with the actual
wolffd@0 30 % frequency of the corresponding chroma.
wolffd@0 31 % c = mirchromagram(...,'Weight',o): specifies the relative radius of
wolffd@0 32 % the weighting window, with respect to the distance between
wolffd@0 33 % frequencies of successive chromas.
wolffd@0 34 % o = 1: each window begins at the centers of the previous one.
wolffd@0 35 % o = .5: each window begins at the end of the previous one.
wolffd@0 36 % (default value)
wolffd@0 37 % mirchromagram(...,'Min',mi) indicates the lowest frequency taken into
wolffd@0 38 % consideration in the spectrum computation, expressed in Hz.
wolffd@0 39 % Default value: 100 Hz. (Gomez, 2006)
wolffd@0 40 % mirchromagram(...,'Max',ma) indicates the highest frequency taken into
wolffd@0 41 % consideration in the spectrum computation, expressed in Hz.
wolffd@0 42 % This upper limit is further shifted to a highest value until
wolffd@0 43 % the frequency range covers an exact multiple of octaves.
wolffd@0 44 % Default value: 5000 Hz. (Gomez, 2006)
wolffd@0 45 % mirchromagram(...,'Res',r) indicates the resolution of the
wolffd@0 46 % chromagram in number of bins per octave.
wolffd@0 47 % Default value, r = 12.
wolffd@0 48 %
wolffd@0 49 % Gómez, E. (2006). Tonal description of music audio signal. Phd thesis,
wolffd@0 50 % Universitat Pompeu Fabra, Barcelona .
wolffd@0 51
wolffd@0 52 cen.key = 'Center';
wolffd@0 53 cen.type = 'Boolean';
wolffd@0 54 cen.default = 0;
wolffd@0 55 option.cen = cen;
wolffd@0 56
wolffd@0 57 nor.key = {'Normal','Norm'};
wolffd@0 58 nor.type = 'Integer';
wolffd@0 59 nor.default = Inf;
wolffd@0 60 option.nor = nor;
wolffd@0 61
wolffd@0 62 wth.key = 'Weight';
wolffd@0 63 wth.type = 'Integer';
wolffd@0 64 wth.default = .5;
wolffd@0 65 option.wth = wth;
wolffd@0 66
wolffd@0 67 tri.key = 'Triangle';
wolffd@0 68 tri.type = 'Boolean';
wolffd@0 69 tri.default = 0;
wolffd@0 70 option.tri = tri;
wolffd@0 71
wolffd@0 72 wrp.key = 'Wrap';
wolffd@0 73 wrp.type = 'Boolean';
wolffd@0 74 wrp.default = 1;
wolffd@0 75 option.wrp = wrp;
wolffd@0 76
wolffd@0 77 plabel.key = 'Pitch';
wolffd@0 78 plabel.type = 'Boolean';
wolffd@0 79 plabel.default = 1;
wolffd@0 80 option.plabel = plabel;
wolffd@0 81
wolffd@0 82 thr.key = {'Threshold','dB'};
wolffd@0 83 thr.type = 'Integer';
wolffd@0 84 thr.default = 20;
wolffd@0 85 option.thr = thr;
wolffd@0 86
wolffd@0 87 min.key = 'Min';
wolffd@0 88 min.type = 'Integer';
wolffd@0 89 min.default = 100;
wolffd@0 90 option.min = min;
wolffd@0 91
wolffd@0 92 max.key = 'Max';
wolffd@0 93 max.type = 'Integer';
wolffd@0 94 max.default = 5000;
wolffd@0 95 option.max = max;
wolffd@0 96
wolffd@0 97 res.key = 'Res';
wolffd@0 98 res.type = 'Integer';
wolffd@0 99 res.default = 12;
wolffd@0 100 option.res = res;
wolffd@0 101
wolffd@0 102 origin.key = 'Tuning';
wolffd@0 103 origin.type = 'Integer';
wolffd@0 104 origin.default = 261.6256;
wolffd@0 105 option.origin = origin;
wolffd@0 106
wolffd@0 107 specif.option = option;
wolffd@0 108 specif.defaultframelength = .2;
wolffd@0 109 specif.defaultframehop = .05;
wolffd@0 110
wolffd@0 111 varargout = mirfunction(@mirchromagram,orig,varargin,nargout,specif,@init,@main);
wolffd@0 112
wolffd@0 113
wolffd@0 114 function [x type] = init(x,option)
wolffd@0 115 if isamir(x,'mirtemporal') || isamir(x,'mirspectrum')
wolffd@0 116 freqmin = option.min;
wolffd@0 117 freqmax = freqmin*2;
wolffd@0 118 while freqmax < option.max
wolffd@0 119 freqmax = freqmax*2;
wolffd@0 120 end
wolffd@0 121 %freqres = freqmin*(2.^(1/option.res)-1);
wolffd@0 122 % Minimal frequency resolution should correspond to frequency range
wolffd@0 123 % between the first two bins of the chromagram
wolffd@0 124
wolffd@0 125 x = mirspectrum(x,'dB',option.thr,'Min',freqmin,'Max',freqmax,...
wolffd@0 126 'NormalInput','MinRes',option.res,'OctaveRatio',.85);
wolffd@0 127 %freqres*.5,...
wolffd@0 128 % 'WarningRes',freqres);
wolffd@0 129 end
wolffd@0 130 type = 'mirchromagram';
wolffd@0 131
wolffd@0 132
wolffd@0 133 function c = main(orig,option,postoption)
wolffd@0 134 if iscell(orig)
wolffd@0 135 orig = orig{1};
wolffd@0 136 end
wolffd@0 137 if option.res == 12
wolffd@0 138 chromascale = {'C','C#','D','D#','E','F','F#','G','G#','A','A#','B'};
wolffd@0 139 else
wolffd@0 140 chromascale = 1:option.res;
wolffd@0 141 option.plabel = 0;
wolffd@0 142 end
wolffd@0 143 if isa(orig,'mirchromagram')
wolffd@0 144 c = modif(orig,option,chromascale);
wolffd@0 145 else
wolffd@0 146 c.plabel = 1;
wolffd@0 147 c.wrap = 0;
wolffd@0 148 c.chromaclass = {};
wolffd@0 149 c.chromafreq = {};
wolffd@0 150 c.register = {};
wolffd@0 151 c = class(c,'mirchromagram',mirdata(orig));
wolffd@0 152 c = purgedata(c);
wolffd@0 153 c = set(c,'Title','Chromagram','Ord','magnitude','Interpolable',0);
wolffd@0 154 if option.wrp
wolffd@0 155 c = set(c,'Abs','chroma class');
wolffd@0 156 else
wolffd@0 157 c = set(c,'Abs','chroma');
wolffd@0 158 end
wolffd@0 159 m = get(orig,'Magnitude');
wolffd@0 160 f = get(orig,'Frequency');
wolffd@0 161 %disp('Computing chromagram...')
wolffd@0 162 fs = get(orig,'Sampling');
wolffd@0 163 n = cell(1,length(m)); % The final structured list of magnitudes.
wolffd@0 164 cc = cell(1,length(m)); % The final structured list of chroma classes.
wolffd@0 165 o = cell(1,length(m)); % The final structured list of octave registers.
wolffd@0 166 p = cell(1,length(m)); % The final structured list of chromas.
wolffd@0 167 cf = cell(1,length(m)); % The final structured list of central frequencies related to chromas.
wolffd@0 168 for i = 1:length(m)
wolffd@0 169 mi = m{i};
wolffd@0 170 fi = f{i};
wolffd@0 171 if not(iscell(mi))
wolffd@0 172 mi = {mi};
wolffd@0 173 fi = {fi};
wolffd@0 174 end
wolffd@0 175 ni = cell(1,length(mi)); % The list of magnitudes.
wolffd@0 176 ci = cell(1,length(mi)); % The list of chroma classes.
wolffd@0 177 oi = cell(1,length(mi)); % The list of octave registers.
wolffd@0 178 pi = cell(1,length(mi)); % The list of absolute chromas.
wolffd@0 179 cfi = cell(1,length(mi)); % The central frequency of each chroma.
wolffd@0 180 for j = 1:length(mi)
wolffd@0 181 mj = mi{j};
wolffd@0 182 fj = fi{j};
wolffd@0 183
wolffd@0 184 % Let's remove the frequencies exceeding the last whole octave.
wolffd@0 185 minfj = min(min(min(fj)));
wolffd@0 186 maxfj = max(max(max(fj)));
wolffd@0 187 maxfj = minfj*2^(floor(log2(maxfj/minfj)));
wolffd@0 188 fz = find(fj(:,1,1,1) > maxfj);
wolffd@0 189 mj(fz,:,:,:) = [];
wolffd@0 190 fj(fz,:,:,:) = [];
wolffd@0 191
wolffd@0 192 [s1 s2 s3] = size(mj);
wolffd@0 193
wolffd@0 194 cj = freq2chro(fj,option.res,option.origin);
wolffd@0 195 if not(ismember(min(cj)+1,cj))
wolffd@0 196 warning('WARNING IN MIRCHROMAGRAM: Frequency resolution of the spectrum is too low.');
wolffd@0 197 display('The conversion of low frequencies into chromas may be incorrect.');
wolffd@0 198 end
wolffd@0 199 ccj = min(min(min(cj))):max(max(max(cj)));
wolffd@0 200 sc = length(ccj); % The size of range of absolute chromas.
wolffd@0 201 mat = zeros(s1,sc);
wolffd@0 202 fc = chro2freq(ccj,option.res,option.origin); % The absolute chromas in Hz.
wolffd@0 203 fl = chro2freq(ccj-1,option.res,option.origin); % Each previous chromas in Hz.
wolffd@0 204 fr = chro2freq(ccj+1,option.res,option.origin); % Each related next chromas in Hz.
wolffd@0 205 for k = 1:sc
wolffd@0 206 rad = find(and(fj(:,1) > fc(k)-option.wth*(fc(k)-fl(k)),...
wolffd@0 207 fj(:,1) < fc(k)-option.wth*(fc(k)-fr(k))));
wolffd@0 208 if option.tri
wolffd@0 209 dist = fc(k) - fj(:,1,1,1);
wolffd@0 210 rad1 = dist/(fc(k) - fl(k))/option.wth;
wolffd@0 211 rad2 = dist/(fc(k) - fr(k))/option.wth;
wolffd@0 212 ndist = max(rad1,rad2);
wolffd@0 213 mat(:,k) = max(min(1-ndist,1),0)/length(rad);
wolffd@0 214 else
wolffd@0 215 mat(rad,k) = ones(length(rad),1)/length(rad);
wolffd@0 216 end
wolffd@0 217 if k ==1 || k == sc
wolffd@0 218 mat(:,k) = mat(:,k)/2;
wolffd@0 219 end
wolffd@0 220 end
wolffd@0 221 nj = zeros(sc,s2,s3);
wolffd@0 222 for k = 1:s2
wolffd@0 223 for l = 1:s3
wolffd@0 224 nj(:,k,l) = (mj(:,k,l)'*mat)';
wolffd@0 225 end
wolffd@0 226 end
wolffd@0 227 cj = mod(ccj',option.res);
wolffd@0 228 oi{j} = floor(ccj/option.res)+4;
wolffd@0 229 if option.plabel
wolffd@0 230 pj = strcat(chromascale(cj+1)',num2str(oi{j}'));
wolffd@0 231 else
wolffd@0 232 pj = ccj'+60;
wolffd@0 233 end
wolffd@0 234 ci{j} = repmat(cj,[1,s2,s3]);
wolffd@0 235 pi{j} = repmat(pj,[1,s2,s3]);
wolffd@0 236 ni{j} = nj;
wolffd@0 237 cfi{j} = fc;
wolffd@0 238 end
wolffd@0 239 n{i} = ni;
wolffd@0 240 cc{i} = ci;
wolffd@0 241 o{i} = oi;
wolffd@0 242 p{i} = pi;
wolffd@0 243 cf{i} = cfi;
wolffd@0 244 end
wolffd@0 245 c = set(c,'Magnitude',n,'Chroma',p,'ChromaClass',cc,...
wolffd@0 246 'ChromaFreq',cf,'Register',o);
wolffd@0 247 c = modif(c,option,chromascale);
wolffd@0 248 c = {c orig};
wolffd@0 249 end
wolffd@0 250
wolffd@0 251
wolffd@0 252 function c = modif(c,option,chromascale)
wolffd@0 253 if option.plabel
wolffd@0 254 c = set(c,'PitchLabel',1);
wolffd@0 255 end
wolffd@0 256 if option.cen || option.nor || option.wrp
wolffd@0 257 n = get(c,'Magnitude');
wolffd@0 258 p = get(c,'Chroma');
wolffd@0 259 cl = get(c,'ChromaClass');
wolffd@0 260 fp = get(c,'FramePos');
wolffd@0 261 n2 = cell(1,length(n));
wolffd@0 262 p2 = cell(1,length(n));
wolffd@0 263 wrp = option.wrp && not(get(c,'Wrap'));
wolffd@0 264 for i = 1:length(n)
wolffd@0 265 ni = n{i};
wolffd@0 266 pi = p{i};
wolffd@0 267 cli = cl{i};
wolffd@0 268 if not(iscell(ni))
wolffd@0 269 ni = {ni};
wolffd@0 270 pi = {pi};
wolffd@0 271 cli = {cli};
wolffd@0 272 end
wolffd@0 273 if wrp
wolffd@0 274 c = set(c,'Wrap',option.wrp);
wolffd@0 275 end
wolffd@0 276 n2i = cell(1,length(ni));
wolffd@0 277 p2i = cell(1,length(ni));
wolffd@0 278 for j = 1:length(ni)
wolffd@0 279 nj = ni{j};
wolffd@0 280 pj = pi{j};
wolffd@0 281 clj = cli{j};
wolffd@0 282 if wrp
wolffd@0 283 n2j = zeros(option.res,size(nj,2),size(nj,3));
wolffd@0 284 for k = 1:size(pj,1)
wolffd@0 285 n2j(clj(k)+1,:,:) = n2j(clj(k)+1,:,:) + nj(k,:,:); % squared sum (parameter)
wolffd@0 286 end
wolffd@0 287 p2i{j} = chromascale';
wolffd@0 288 else
wolffd@0 289 n2j = nj;
wolffd@0 290 p2i{j} = pi{j};
wolffd@0 291 end
wolffd@0 292 if option.cen
wolffd@0 293 n2j = n2j - repmat(mean(n2j),[size(n2j,1),1,1]);
wolffd@0 294 end
wolffd@0 295 if option.nor
wolffd@0 296 n2j = n2j ./ repmat(vectnorm(n2j,option.nor) + ...
wolffd@0 297 repmat(1e-6,[1,size(n2j,2),size(n2j,3)] )...
wolffd@0 298 ,[size(n2j,1),1,1]);
wolffd@0 299 end
wolffd@0 300 n2i{j} = n2j;
wolffd@0 301 end
wolffd@0 302 n2{i} = n2i;
wolffd@0 303 p2{i} = p2i;
wolffd@0 304 end
wolffd@0 305 c = set(c,'Magnitude',n2,'Chroma',p2,'FramePos',fp);
wolffd@0 306 end
wolffd@0 307
wolffd@0 308
wolffd@0 309 function c = freq2chro(f,res,origin)
wolffd@0 310 c = round(res*log2(f/origin));
wolffd@0 311
wolffd@0 312
wolffd@0 313 function f = chro2freq(c,res,origin)
wolffd@0 314 f = 2.^(c/res)*origin;
wolffd@0 315
wolffd@0 316
wolffd@0 317 function y = vectnorm(x,p)
wolffd@0 318 if isinf(p)
wolffd@0 319 y = max(x);
wolffd@0 320 else
wolffd@0 321 y = sum(abs(x).^p).^(1/p);
wolffd@0 322 end