Mercurial > hg > camir-aes2014

annotate toolboxes/MIRtoolbox1.3.2/MIRToolbox/mirtempo.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 = mirtempo(x,varargin)
wolffd@0 2 % t = mirtempo(x) evaluates the tempo in beats per minute (BPM).
wolffd@0 3 % Optional arguments:
wolffd@0 4 % mirtempo(...,'Total',m) selects not only the best tempo, but the m
wolffd@0 5 % best tempos.
wolffd@0 6 % mirtempo(...,'Frame',l,h) orders a frame decomposition of window
wolffd@0 7 % length l (in seconds) and hop factor h, expressed relatively to
wolffd@0 8 % the window length. For instance h = 1 indicates no overlap.
wolffd@0 9 % Default values: l = 3 seconds and h = .1
wolffd@0 10 % mirtempo(...,'Min',mi) indicates the lowest tempo taken into
wolffd@0 11 % consideration, expressed in bpm.
wolffd@0 12 % Default value: 40 bpm.
wolffd@0 13 % mirtempo(...,'Max',ma) indicates the highest tempo taken into
wolffd@0 14 % consideration, expressed in bpm.
wolffd@0 15 % Default value: 200 bpm.
wolffd@0 16 % mirtempo(...,s) selects the tempo estimation strategy:
wolffd@0 17 % s = 'Autocor': Approach based on the computation of the
wolffd@0 18 % autocorrelation. (Default strategy)
wolffd@0 19 % Option associated to the mirautocor function can be
wolffd@0 20 % passed here as well (see help mirautocor):
wolffd@0 21 % 'Enhanced' (toggled on by default here)
wolffd@0 22 % s = 'Spectrum': Approach based on the computation of the
wolffd@0 23 % spectrum .
wolffd@0 24 % Option associated to the mirspectrum function can be
wolffd@0 25 % passed here as well (see help mirspectrum):
wolffd@0 26 % 'ZeroPad' (set by default to 10000 samples)
wolffd@0 27 % 'Prod' (toggled off by default)
wolffd@0 28 % These two strategies can be combined: the autocorrelation
wolffd@0 29 % function is translated into the frequency domain in order
wolffd@0 30 % to be compared to the spectrum curve.
wolffd@0 31 % tempo(...,'Autocor','Spectrum') multiplies the two curves.
wolffd@0 32 % Alternatively, an autocorrelation function ac or a spectrum sp
wolffd@0 33 % can be directly passed to the function tempo:
wolffd@0 34 % mirtempo(ac) or mirtempo(sp)
wolffd@0 35 % The options related to the onset detection phase can be specified
wolffd@0 36 % here as well (see help mironsets):
wolffd@0 37 % onset detection strategies: 'Envelope', 'DiffEnvelope'
wolffd@0 38 % (corresponding to 'Envelope', 'Diff'), 'SpectralFlux,
wolffd@0 39 % 'Pitch', 'Log', 'Mu', 'Filterbank'
wolffd@0 40 % mironsets(...,'Sum',w) specifies when to sum the channels.
wolffd@0 41 % Possible values:
wolffd@0 42 % w = 'Before': sum before the autocorrelation or
wolffd@0 43 % spectrum computation.
wolffd@0 44 % w = 'After': autocorrelation or spectrum computed
wolffd@0 45 % for each band, and summed into a "summary".
wolffd@0 46 % mirenvelope options: 'HalfwaveCenter','Diff' (toggled on by
wolffd@0 47 % default here),'HalfwaveDiff','Center','Smooth',
wolffd@0 48 % 'Sampling'
wolffd@0 49 % mirflux options: 'Inc','Halfwave','Complex','Median'
wolffd@0 50 % mirtempo(...,'Resonance',r) specifies the resonance curve, which
wolffd@0 51 % emphasizes the periods that are more easily perceived.
wolffd@0 52 % Possible values: 'ToiviainenSnyder' (default), 0 (toggled off)
wolffd@0 53 % Optional arguments used for the peak picking (cf. help mirpeaks)
wolffd@0 54 % mirtempo(...,'Contrast',thr): a threshold value. A given local
wolffd@0 55 % maximum will be considered as a peak if its distance with the
wolffd@0 56 % previous and successive local minima (if any) is higher than
wolffd@0 57 % this threshold. This distance is expressed with respect to the
wolffd@0 58 % total amplitude of the autocorrelation function.
wolffd@0 59 % if no value for thr is given, the value thr=0.1 is chosen
wolffd@0 60 % by default.
wolffd@0 61 %
wolffd@0 62 % [t,p] = mirtempo(...) also displays the result of the signal analysis
wolffd@0 63 % leading to the tempo estimation, and shows in particular the
wolffd@0 64 % peaks corresponding to the tempo values.
wolffd@0 65
wolffd@0 66
wolffd@0 67 sum.key = 'Sum';
wolffd@0 68 sum.type = 'String';
wolffd@0 69 sum.choice = {'Before','After','Adjacent',0};
wolffd@0 70 sum.default = 'Before';
wolffd@0 71 option.sum = sum;
wolffd@0 72
wolffd@0 73 %% options related to mironsets:
wolffd@0 74
wolffd@0 75 frame.key = 'Frame';
wolffd@0 76 frame.type = 'Integer';
wolffd@0 77 frame.number = 2;
wolffd@0 78 frame.default = [0 0];
wolffd@0 79 frame.keydefault = [3 .1];
wolffd@0 80 option.frame = frame;
wolffd@0 81
wolffd@0 82 fea.type = 'String';
wolffd@0 83 fea.choice = {'Envelope','DiffEnvelope','SpectralFlux','Pitch'};
wolffd@0 84 fea.default = 'Envelope';
wolffd@0 85 option.fea = fea;
wolffd@0 86
wolffd@0 87 %% options related to 'Envelope':
wolffd@0 88
wolffd@0 89 envmeth.key = 'Method';
wolffd@0 90 envmeth.type = 'String';
wolffd@0 91 envmeth.choice = {'Filter','Spectro'};
wolffd@0 92 envmeth.default = 'Filter';
wolffd@0 93 option.envmeth = envmeth;
wolffd@0 94
wolffd@0 95 %% options related to 'Filter':
wolffd@0 96
wolffd@0 97 fb.key = 'Filterbank';
wolffd@0 98 fb.type = 'Integer';
wolffd@0 99 fb.default = 10;
wolffd@0 100 option.fb = fb;
wolffd@0 101
wolffd@0 102 fbtype.key = 'FilterbankType';
wolffd@0 103 fbtype.type = 'String';
wolffd@0 104 fbtype.choice = {'Gammatone','Scheirer','Klapuri'};
wolffd@0 105 fbtype.default = 'Gammatone';
wolffd@0 106 option.fbtype = fbtype;
wolffd@0 107
wolffd@0 108 ftype.key = 'FilterType';
wolffd@0 109 ftype.type = 'String';
wolffd@0 110 ftype.choice = {'IIR','HalfHann'};
wolffd@0 111 ftype.default = 'IIR';
wolffd@0 112 option.ftype = ftype;
wolffd@0 113
wolffd@0 114 %% options related to 'Spectro':
wolffd@0 115
wolffd@0 116 band.type = 'String';
wolffd@0 117 band.choice = {'Freq','Mel','Bark','Cents'};
wolffd@0 118 band.default = 'Freq';
wolffd@0 119 option.band = band;
wolffd@0 120
wolffd@0 121
wolffd@0 122 chwr.key = 'HalfwaveCenter';
wolffd@0 123 chwr.type = 'Boolean';
wolffd@0 124 chwr.default = 0;
wolffd@0 125 option.chwr = chwr;
wolffd@0 126
wolffd@0 127 diff.key = 'Diff';
wolffd@0 128 diff.type = 'Boolean';
wolffd@0 129 diff.default = 1; % Different default for mirtempo
wolffd@0 130 option.diff = diff;
wolffd@0 131
wolffd@0 132 diffhwr.key = 'HalfwaveDiff';
wolffd@0 133 diffhwr.type = 'Integer';
wolffd@0 134 diffhwr.default = 0;
wolffd@0 135 diffhwr.keydefault = 1;
wolffd@0 136 option.diffhwr = diffhwr;
wolffd@0 137
wolffd@0 138 lambda.key = 'Lambda';
wolffd@0 139 lambda.type = 'Integer';
wolffd@0 140 lambda.default = 1;
wolffd@0 141 option.lambda = lambda;
wolffd@0 142
wolffd@0 143 mu.key = 'Mu';
wolffd@0 144 mu.type = 'Boolean';
wolffd@0 145 mu.default = 0;
wolffd@0 146 option.mu = mu;
wolffd@0 147
wolffd@0 148 log.key = 'Log';
wolffd@0 149 log.type = 'Boolean';
wolffd@0 150 log.default = 0;
wolffd@0 151 option.log = log;
wolffd@0 152
wolffd@0 153 c.key = 'Center';
wolffd@0 154 c.type = 'Boolean';
wolffd@0 155 c.default = 0;
wolffd@0 156 option.c = c;
wolffd@0 157
wolffd@0 158 aver.key = 'Smooth';
wolffd@0 159 aver.type = 'Integer';
wolffd@0 160 aver.default = 0;
wolffd@0 161 aver.keydefault = 30;
wolffd@0 162 option.aver = aver;
wolffd@0 163
wolffd@0 164 sampling.key = 'Sampling';
wolffd@0 165 sampling.type = 'Integer';
wolffd@0 166 sampling.default = 0;
wolffd@0 167 option.sampling = sampling;
wolffd@0 168
wolffd@0 169 %% options related to 'SpectralFlux'
wolffd@0 170
wolffd@0 171 complex.key = 'Complex';
wolffd@0 172 complex.type = 'Boolean';
wolffd@0 173 complex.default = 0;
wolffd@0 174 option.complex = complex;
wolffd@0 175
wolffd@0 176 inc.key = 'Inc';
wolffd@0 177 inc.type = 'Boolean';
wolffd@0 178 inc.default = 1;
wolffd@0 179 option.inc = inc;
wolffd@0 180
wolffd@0 181 median.key = 'Median';
wolffd@0 182 median.type = 'Integer';
wolffd@0 183 median.number = 2;
wolffd@0 184 median.default = [.2 1.3];
wolffd@0 185 option.median = median;
wolffd@0 186
wolffd@0 187 hw.key = 'Halfwave';
wolffd@0 188 hw.type = 'Boolean';
wolffd@0 189 hw.default = 1;
wolffd@0 190 option.hw = hw;
wolffd@0 191
wolffd@0 192
wolffd@0 193 %% options related to mirautocor:
wolffd@0 194
wolffd@0 195 aut.key = 'Autocor';
wolffd@0 196 aut.type = 'Integer';
wolffd@0 197 aut.default = 0;
wolffd@0 198 aut.keydefault = 1;
wolffd@0 199 option.aut = aut;
wolffd@0 200
wolffd@0 201 nw.key = 'NormalWindow';
wolffd@0 202 nw.default = 0;
wolffd@0 203 option.nw = nw;
wolffd@0 204
wolffd@0 205 enh.key = 'Enhanced';
wolffd@0 206 enh.type = 'Integers';
wolffd@0 207 enh.default = 2:10;
wolffd@0 208 enh.keydefault = 2:10;
wolffd@0 209 option.enh = enh;
wolffd@0 210
wolffd@0 211 r.key = 'Resonance';
wolffd@0 212 r.type = 'String';
wolffd@0 213 r.choice = {'ToiviainenSnyder','vanNoorden',0,'off','no'};
wolffd@0 214 r.default = 'ToiviainenSnyder';
wolffd@0 215 option.r = r;
wolffd@0 216
wolffd@0 217
wolffd@0 218 %% options related to mirspectrum:
wolffd@0 219
wolffd@0 220 spe.key = 'Spectrum';
wolffd@0 221 spe.type = 'Integer';
wolffd@0 222 spe.default = 0;
wolffd@0 223 spe.keydefault = 1;
wolffd@0 224 option.spe = spe;
wolffd@0 225
wolffd@0 226 zp.key = 'ZeroPad';
wolffd@0 227 zp.type = 'Integer';
wolffd@0 228 zp.default = 10000;
wolffd@0 229 zp.keydefault = Inf;
wolffd@0 230 option.zp = zp;
wolffd@0 231
wolffd@0 232 prod.key = 'Prod';
wolffd@0 233 prod.type = 'Integers';
wolffd@0 234 prod.default = 0;
wolffd@0 235 prod.keydefault = 2:6;
wolffd@0 236 option.prod = prod;
wolffd@0 237
wolffd@0 238
wolffd@0 239 %% options related to the peak detection
wolffd@0 240
wolffd@0 241 m.key = 'Total';
wolffd@0 242 m.type = 'Integer';
wolffd@0 243 m.default = 1;
wolffd@0 244 option.m = m;
wolffd@0 245
wolffd@0 246 thr.key = 'Contrast';
wolffd@0 247 thr.type = 'Integer';
wolffd@0 248 thr.default = 0.1;
wolffd@0 249 option.thr = thr;
wolffd@0 250
wolffd@0 251 mi.key = 'Min';
wolffd@0 252 mi.type = 'Integer';
wolffd@0 253 mi.default = 40;
wolffd@0 254 option.mi = mi;
wolffd@0 255
wolffd@0 256 ma.key = 'Max';
wolffd@0 257 ma.type = 'Integer';
wolffd@0 258 ma.default = 200;
wolffd@0 259 option.ma = ma;
wolffd@0 260
wolffd@0 261 track.key = 'Track';
wolffd@0 262 track.type = 'Boolean';
wolffd@0 263 track.default = 0;
wolffd@0 264 option.track = track;
wolffd@0 265
wolffd@0 266 pref.key = 'Pref';
wolffd@0 267 pref.type = 'Integer';
wolffd@0 268 pref.number = 2;
wolffd@0 269 pref.default = [0 .2];
wolffd@0 270 option.pref = pref;
wolffd@0 271
wolffd@0 272 perio.key = 'Periodicity';
wolffd@0 273 perio.type = 'Boolean';
wolffd@0 274 perio.default = 0;
wolffd@0 275 option.perio = perio;
wolffd@0 276
wolffd@0 277 specif.option = option;
wolffd@0 278
wolffd@0 279 varargout = mirfunction(@mirtempo,x,varargin,nargout,specif,@init,@main);
wolffd@0 280
wolffd@0 281
wolffd@0 282 %% INIT
wolffd@0 283
wolffd@0 284 function [y type] = init(x,option)
wolffd@0 285 if iscell(x)
wolffd@0 286 x = x{1};
wolffd@0 287 end
wolffd@0 288 if option.perio
wolffd@0 289 option.m = 3;
wolffd@0 290 option.enh = 2:10;
wolffd@0 291 end
wolffd@0 292 if not(isamir(x,'mirautocor')) && not(isamir(x,'mirspectrum'))
wolffd@0 293 if isframed(x) && strcmpi(option.fea,'Envelope') && not(isamir(x,'mirscalar'))
wolffd@0 294 warning('WARNING IN MIRTEMPO: The input should not be already decomposed into frames.');
wolffd@0 295 disp(['Suggestion: Use the ''Frame'' option instead.'])
wolffd@0 296 end
wolffd@0 297 if strcmpi(option.sum,'Before')
wolffd@0 298 optionsum = 1;
wolffd@0 299 elseif strcmpi(option.sum,'Adjacent')
wolffd@0 300 optionsum = 5;
wolffd@0 301 else
wolffd@0 302 optionsum = 0;
wolffd@0 303 end
wolffd@0 304 if option.frame.length.val
wolffd@0 305 x = mironsets(x,option.fea,'Filterbank',option.fb,...
wolffd@0 306 'FilterbankType',option.fbtype,...
wolffd@0 307 'FilterType',option.ftype,...
wolffd@0 308 'Sum',optionsum,'Method',option.envmeth,...
wolffd@0 309 option.band,'Center',option.c,...
wolffd@0 310 'HalfwaveCenter',option.chwr,'Diff',option.diff,...
wolffd@0 311 'HalfwaveDiff',option.diffhwr,'Lambda',option.lambda,...
wolffd@0 312 'Smooth',option.aver,'Sampling',option.sampling,...
wolffd@0 313 'Complex',option.complex,'Inc',option.inc,...
wolffd@0 314 'Median',option.median(1),option.median(2),...
wolffd@0 315 'Halfwave',option.hw,'Detect',0,...
wolffd@0 316 'Mu',option.mu,'Log',option.log,...
wolffd@0 317 'Frame',option.frame.length.val,...
wolffd@0 318 option.frame.length.unit,...
wolffd@0 319 option.frame.hop.val,...
wolffd@0 320 option.frame.hop.unit);
wolffd@0 321 else
wolffd@0 322 x = mironsets(x,option.fea,'Filterbank',option.fb,...
wolffd@0 323 'FilterbankType',option.fbtype,...
wolffd@0 324 'FilterType',option.ftype,...
wolffd@0 325 'Sum',optionsum,'Method',option.envmeth,...
wolffd@0 326 option.band,'Center',option.c,...
wolffd@0 327 'HalfwaveCenter',option.chwr,'Diff',option.diff,...
wolffd@0 328 'HalfwaveDiff',option.diffhwr,'Lambda',option.lambda,...
wolffd@0 329 'Smooth',option.aver,'Sampling',option.sampling,...
wolffd@0 330 'Complex',option.complex,'Inc',option.inc,...
wolffd@0 331 'Median',option.median(1),option.median(2),...
wolffd@0 332 'Halfwave',option.hw,'Detect',0,...
wolffd@0 333 'Mu',option.mu,'Log',option.log);
wolffd@0 334 end
wolffd@0 335 end
wolffd@0 336 if option.aut == 0 && option.spe == 0
wolffd@0 337 option.aut = 1;
wolffd@0 338 end
wolffd@0 339 if isamir(x,'mirautocor') || (option.aut && not(option.spe))
wolffd@0 340 y = mirautocor(x,'Min',60/option.ma,'Max',60/option.mi,...
wolffd@0 341 'Enhanced',option.enh,...'NormalInput','coeff',...
wolffd@0 342 'Resonance',option.r,'NormalWindow',option.nw);
wolffd@0 343 elseif isamir(x,'mirspectrum') || (option.spe && not(option.aut))
wolffd@0 344 y = mirspectrum(x,'Min',option.mi/60,'Max',option.ma/60,...
wolffd@0 345 'Prod',option.prod,...'NormalInput',...
wolffd@0 346 'ZeroPad',option.zp,'Resonance',option.r);
wolffd@0 347 elseif option.spe && option.aut
wolffd@0 348 ac = mirautocor(x,'Min',60/option.ma,'Max',60/option.mi,...
wolffd@0 349 'Enhanced',option.enh,...'NormalInput','coeff',...
wolffd@0 350 'Resonance',option.r);
wolffd@0 351 sp = mirspectrum(x,'Min',option.mi/60,'Max',option.ma/60,...
wolffd@0 352 'Prod',option.prod,...'NormalInput',...
wolffd@0 353 'ZeroPad',option.zp,'Resonance',option.r);
wolffd@0 354 y = ac*sp;
wolffd@0 355 end
wolffd@0 356 if ischar(option.sum)
wolffd@0 357 y = mirsum(y);
wolffd@0 358 end
wolffd@0 359 y = mirpeaks(y,'Total',option.m,'Track',option.track,...
wolffd@0 360 'Pref',option.pref(1),option.pref(2),...
wolffd@0 361 'Contrast',option.thr,'NoBegin','NoEnd',...
wolffd@0 362 'Normalize','Local');
wolffd@0 363 type = {'mirscalar',mirtype(y)};
wolffd@0 364
wolffd@0 365
wolffd@0 366 %% MAIN
wolffd@0 367
wolffd@0 368 function o = main(p,option,postoption)
wolffd@0 369 if iscell(p)
wolffd@0 370 p = p{1};
wolffd@0 371 end
wolffd@0 372 pt = get(p,'TrackPrecisePos');
wolffd@0 373 track = 1;
wolffd@0 374 if isempty(pt) || isempty(pt{1})
wolffd@0 375 pt = get(p,'PeakPrecisePos');
wolffd@0 376 track = 0;
wolffd@0 377 end
wolffd@0 378 bpm = cell(1,length(pt));
wolffd@0 379 for j = 1:length(pt)
wolffd@0 380 bpm{j} = cell(1,length(pt{j}));
wolffd@0 381 for k = 1:length(pt{j})
wolffd@0 382 ptk = pt{j}{k};
wolffd@0 383 bpmk = cell(1,size(ptk,2));
wolffd@0 384 for h = 1:size(ptk,3)
wolffd@0 385 for l = 1:size(ptk,2)
wolffd@0 386 ptl = ptk{1,l,h};
wolffd@0 387 if isempty(ptl)
wolffd@0 388 bpmk{1,l,h} = NaN;
wolffd@0 389 else
wolffd@0 390 if isa(p,'mirautocor') && not(get(p,'FreqDomain'))
wolffd@0 391 bpmk{1,l,h} = 60./ptl;
wolffd@0 392 else
wolffd@0 393 bpmk{1,l,h} = ptl*60;
wolffd@0 394 end
wolffd@0 395 end
wolffd@0 396 end
wolffd@0 397 end
wolffd@0 398 if track
wolffd@0 399 bpmk = bpmk{1};
wolffd@0 400 end
wolffd@0 401 bpm{j}{k} = bpmk;
wolffd@0 402 end
wolffd@0 403 end
wolffd@0 404 t = mirscalar(p,'Data',bpm,'Title','Tempo','Unit','bpm');
wolffd@0 405 o = {t,p};