Mercurial > hg > camir-aes2014

annotate toolboxes/MIRtoolbox1.3.2/MIRToolbox/@miraudio/miraudio.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 = miraudio(orig,varargin)
wolffd@0 2 % a = miraudio('filename') loads the sound file 'filename' (in WAV or AU
wolffd@0 3 % format) into a miraudio object.
wolffd@0 4 % a = miraudio('Folder') loads all the sound files in the CURRENT folder
wolffd@0 5 % into a miraudio object.
wolffd@0 6 % a = miraudio(v,sr), where v is a column vector, translates the vector v
wolffd@0 7 % into a miraudio object. The sampling frequency is set to sr Hertz.
wolffd@0 8 % Default value for sr: 44100 Hz.
wolffd@0 9 % a = miraudio(b, ...), where b is already a miraudio object, performs
wolffd@0 10 % operations on b specified by the optional arguments (see below).
wolffd@0 11 %
wolffd@0 12 % Transformation options:
wolffd@0 13 % miraudio(...,'Mono',0) does not perform the default summing of
wolffd@0 14 % channels into one single mono track, but instead stores each
wolffd@0 15 % channel of the initial soundfile separately.
wolffd@0 16 % miraudio(...,'Center') centers the signals.
wolffd@0 17 % miraudio(...,'Sampling',r) resamples at sampling rate r (in Hz).
wolffd@0 18 % (Requires the Signal Processing Toolbox.)
wolffd@0 19 % miraudio(...,'Normal') normalizes with respect to RMS energy.
wolffd@0 20 % Extraction options:
wolffd@0 21 % miraudio(...,'Extract',t1,t2,u,f) extracts the signal between dates
wolffd@0 22 % t1 and t2, expressed in the unit u.
wolffd@0 23 % Possible values for u:
wolffd@0 24 % 's' (seconds, by default),
wolffd@0 25 % 'sp' (sample index, starting from 1).
wolffd@0 26 % The additional optional argument f indicates the referential
wolffd@0 27 % origin of the temporal positions. Possible values for f:
wolffd@0 28 % 'Start' (by default)
wolffd@0 29 % 'Middle' (of the sequence)
wolffd@0 30 % 'End' of the sequence
wolffd@0 31 % When using 'Middle' or 'End', negative values for t1 or t2
wolffd@0 32 % indicate values before the middle or the end of the audio
wolffd@0 33 % sequence.
wolffd@0 34 % miraudio(...,'Trim') trims the pseudo-silence beginning and end off
wolffd@0 35 % the audio file. Silent frames are frames with RMS below t times
wolffd@0 36 % the medium RMS of the whole audio file.
wolffd@0 37 % Default value: t = 0.06
wolffd@0 38 % instead of 'Trim':
wolffd@0 39 % 'TrimStart' only trims the beginning of the audio file,
wolffd@0 40 % 'TrimEnd' only trims the end.
wolffd@0 41 % miraudio(...,'TrimThreshold',t) specifies the trimming threshold t.
wolffd@0 42 % miraudio(...,'Channel',c) or miraudio(...,'Channels',c) selects the
wolffd@0 43 % channels indicated by the (array of) integer(s) c.
wolffd@0 44 % Labeling option:
wolffd@0 45 % miraudio(...,'Label',l) labels the audio signal(s) following the
wolffd@0 46 % label(s) l.
wolffd@0 47 % If l is a (series of) number(s), the audio signal(s) are
wolffd@0 48 % labelled using the substring of their respective file name of
wolffd@0 49 % index l. If l=0, the audio signal(s) are labelled using the
wolffd@0 50 % whole file name.
wolffd@0 51
wolffd@0 52
wolffd@0 53 if isnumeric(orig)
wolffd@0 54 if size(orig,2) > 1 || size(orig,3) > 1
wolffd@0 55 mirerror('MIRAUDIO','Only column vectors can be imported into mirtoolbox.');
wolffd@0 56 end
wolffd@0 57 if nargin == 1
wolffd@0 58 f = 44100;
wolffd@0 59 else
wolffd@0 60 f = varargin{1};
wolffd@0 61 end
wolffd@0 62 b = 32;
wolffd@0 63 if size(orig,1) == 1
wolffd@0 64 orig = orig';
wolffd@0 65 end
wolffd@0 66 tp = (0:size(orig,1)-1)'/f;
wolffd@0 67 t = mirtemporal([],'Time',{{tp}},'Data',{{orig}},...
wolffd@0 68 'FramePos',{{tp([1 end])}},'Sampling',{f},...
wolffd@0 69 'Name',{inputname(1)},'Label',{{}},'Clusters',{{}},...
wolffd@0 70 'Channels',[],'Centered',0,'NBits',{b},...
wolffd@0 71 'Title','Audio signal',...
wolffd@0 72 'PeakPos',{{{}}},'PeakVal',{{{}}},'PeakMode',{{{}}});
wolffd@0 73 aa.fresh = 1;
wolffd@0 74 varargout = {class(aa,'miraudio',t)};
wolffd@0 75 return
wolffd@0 76 end
wolffd@0 77
wolffd@0 78
wolffd@0 79 center.key = 'Center';
wolffd@0 80 center.type = 'Boolean';
wolffd@0 81 center.default = 0;
wolffd@0 82 center.when = 'After';
wolffd@0 83 option.center = center;
wolffd@0 84
wolffd@0 85 normal.key = 'Normal';
wolffd@0 86 normal.type = 'Boolean';
wolffd@0 87 normal.default = 0;
wolffd@0 88 normal.when = 'After';
wolffd@0 89 option.normal = normal;
wolffd@0 90
wolffd@0 91 extract.key = {'Extract','Excerpt'};
wolffd@0 92 extract.type = 'Integer';
wolffd@0 93 extract.number = 2;
wolffd@0 94 extract.default = [];
wolffd@0 95 extract.unit = {'s','sp'};
wolffd@0 96 extract.defaultunit = 's';
wolffd@0 97 extract.from = {'Start','Middle','End'};
wolffd@0 98 extract.defaultfrom = 'Start';
wolffd@0 99 option.extract = extract;
wolffd@0 100
wolffd@0 101 trim.type = 'String';
wolffd@0 102 trim.choice = {'NoTrim','Trim','TrimBegin','TrimStart','TrimEnd'};
wolffd@0 103 trim.default = 'NoTrim';
wolffd@0 104 trim.when = 'After';
wolffd@0 105 option.trim = trim;
wolffd@0 106
wolffd@0 107 trimthreshold.key = 'TrimThreshold';
wolffd@0 108 trimthreshold.type = 'Integer';
wolffd@0 109 trimthreshold.default = .06;
wolffd@0 110 trimthreshold.when = 'After';
wolffd@0 111 option.trimthreshold = trimthreshold;
wolffd@0 112
wolffd@0 113 label.key = 'Label';
wolffd@0 114 label.default = '';
wolffd@0 115 label.when = 'After';
wolffd@0 116 option.label = label;
wolffd@0 117
wolffd@0 118 sampling.key = 'Sampling';
wolffd@0 119 sampling.type = 'Integer';
wolffd@0 120 sampling.default = 0;
wolffd@0 121 sampling.when = 'Both';
wolffd@0 122 option.sampling = sampling;
wolffd@0 123
wolffd@0 124 % segment.key = 'Segment';
wolffd@0 125 % segment.type = 'Integer';
wolffd@0 126 % segment.default = [];
wolffd@0 127 % segment.when = 'After';
wolffd@0 128 % option.segment = segment;
wolffd@0 129
wolffd@0 130 reverse.key = 'Reverse';
wolffd@0 131 reverse.type = 'Boolean';
wolffd@0 132 reverse.default = 0;
wolffd@0 133 reverse.when = 'After';
wolffd@0 134 option.reverse = reverse;
wolffd@0 135
wolffd@0 136 mono.key = 'Mono';
wolffd@0 137 mono.type = 'Boolean';
wolffd@0 138 mono.default = NaN;
wolffd@0 139 mono.when = 'After';
wolffd@0 140 option.mono = mono;
wolffd@0 141
wolffd@0 142 separate.key = 'SeparateChannels';
wolffd@0 143 separate.type = 'Boolean';
wolffd@0 144 separate.default = 0;
wolffd@0 145 option.separate = separate;
wolffd@0 146
wolffd@0 147 Ch.key = {'Channel','Channels'};
wolffd@0 148 Ch.type = 'Integer';
wolffd@0 149 Ch.default = [];
wolffd@0 150 Ch.when = 'After';
wolffd@0 151 option.Ch = Ch;
wolffd@0 152
wolffd@0 153 specif.option = option;
wolffd@0 154
wolffd@0 155 specif.beforechunk = {@beforechunk,'normal'};
wolffd@0 156 specif.eachchunk = @eachchunk;
wolffd@0 157 specif.combinechunk = @combinechunk;
wolffd@0 158
wolffd@0 159 if nargin > 1 && ischar(varargin{1}) && strcmp(varargin{1},'Now')
wolffd@0 160 if nargin > 2
wolffd@0 161 extract = varargin{2};
wolffd@0 162 else
wolffd@0 163 extract = [];
wolffd@0 164 end
wolffd@0 165 para = [];
wolffd@0 166 varargout = {main(orig,[],para,[],extract)};
wolffd@0 167 else
wolffd@0 168 varargout = mirfunction(@miraudio,orig,varargin,nargout,specif,@init,@main);
wolffd@0 169 end
wolffd@0 170 if isempty(varargout)
wolffd@0 171 varargout = {{}};
wolffd@0 172 end
wolffd@0 173
wolffd@0 174
wolffd@0 175 function [x type] = init(x,option)
wolffd@0 176 if isa(x,'mirdesign')
wolffd@0 177 if option.sampling
wolffd@0 178 x = setresampling(x,option.sampling);
wolffd@0 179 end
wolffd@0 180 end
wolffd@0 181 type = 'miraudio';
wolffd@0 182
wolffd@0 183
wolffd@0 184 function a = main(orig,option,after,index,extract)
wolffd@0 185 if iscell(orig)
wolffd@0 186 orig = orig{1};
wolffd@0 187 end
wolffd@0 188 if ischar(orig)
wolffd@0 189 if nargin < 5
wolffd@0 190 extract = [];
wolffd@0 191 end
wolffd@0 192 [d{1},tp{1},fp{1},f{1},b{1},n{1},ch{1}] = mirread(extract,orig,1,0);
wolffd@0 193 t = mirtemporal([],'Time',tp,'Data',d,'FramePos',fp,'Sampling',f,...
wolffd@0 194 'Name',n,'Label',cell(1,length(d)),...
wolffd@0 195 'Clusters',cell(1,length(d)),...
wolffd@0 196 'Channels',ch,'Centered',0,'NBits',b);
wolffd@0 197 t = set(t,'Title','Audio waveform');
wolffd@0 198 a.fresh = 1;
wolffd@0 199 a = class(a,'miraudio',t);
wolffd@0 200 else
wolffd@0 201 if not(isempty(option)) && not(isempty(option.extract))
wolffd@0 202 if not(isstruct(after))
wolffd@0 203 after = struct;
wolffd@0 204 end
wolffd@0 205 after.extract = option.extract;
wolffd@0 206 end
wolffd@0 207 if isa(orig,'miraudio')
wolffd@0 208 a = orig;
wolffd@0 209 else
wolffd@0 210 a.fresh = 1;
wolffd@0 211 a = class(a,'miraudio',orig);
wolffd@0 212 end
wolffd@0 213 end
wolffd@0 214 if not(isempty(after))
wolffd@0 215 a = post(a,after);
wolffd@0 216 end
wolffd@0 217
wolffd@0 218
wolffd@0 219 function a = post(a,para)
wolffd@0 220 if a.fresh && isfield(para,'mono')
wolffd@0 221 a.fresh = 0;
wolffd@0 222 if isnan(para.mono)
wolffd@0 223 para.mono = 1;
wolffd@0 224 end
wolffd@0 225 end
wolffd@0 226 if isfield(para,'mono') && para.mono == 1
wolffd@0 227 a = mirsum(a,'Mean');
wolffd@0 228 end
wolffd@0 229 d = get(a,'Data');
wolffd@0 230 t = get(a,'Time');
wolffd@0 231 ac = get(a,'AcrossChunks');
wolffd@0 232 f = get(a,'Sampling');
wolffd@0 233 cl = get(a,'Clusters');
wolffd@0 234 for h = 1:length(d)
wolffd@0 235 for k = 1:length(d{h})
wolffd@0 236 tk = t{h}{k};
wolffd@0 237 dk = d{h}{k};
wolffd@0 238 if isfield(para,'extract') && not(isempty(para.extract))
wolffd@0 239 t1 = para.extract(1);
wolffd@0 240 t2 = para.extract(2);
wolffd@0 241 if para.extract(4)
wolffd@0 242 if para.extract(4) == 1
wolffd@0 243 shift = round(size(tk,1)/2);
wolffd@0 244 elseif para.extract(4) == 2
wolffd@0 245 shift = size(tk,1);
wolffd@0 246 end
wolffd@0 247 if para.extract(3)
wolffd@0 248 shift = tk(shift,1,1);
wolffd@0 249 end
wolffd@0 250 t1 = t1+shift;
wolffd@0 251 t2 = t2+shift;
wolffd@0 252 end
wolffd@0 253 if para.extract(3) % in seconds
wolffd@0 254 ft = find(tk>=t1 & tk<=t2);
wolffd@0 255 else % in samples
wolffd@0 256 if not(t1)
wolffd@0 257 warning('WARNING IN MIRAUDIO: Extract sample positions should be real positive integers.')
wolffd@0 258 display('Positions incremented by one.');
wolffd@0 259 t1 = t1+1;
wolffd@0 260 t2 = t2+1;
wolffd@0 261 end
wolffd@0 262 ft = t1:t2;
wolffd@0 263 end
wolffd@0 264 tk = tk(ft,:,:);
wolffd@0 265 dk = dk(ft,:,:);
wolffd@0 266 end
wolffd@0 267 if isfield(para,'Ch') && not(isempty(para.Ch))
wolffd@0 268 dk = dk(:,:,para.Ch);
wolffd@0 269 end
wolffd@0 270 if isfield(para,'center') && para.center
wolffd@0 271 dk = center(dk);
wolffd@0 272 a = set(a,'Centered',1);
wolffd@0 273 end
wolffd@0 274 if isfield(para,'normal') && para.normal
wolffd@0 275 nl = size(dk,1);
wolffd@0 276 nc = size(dk,3);
wolffd@0 277 if isempty(ac)
wolffd@0 278 ee = 0;
wolffd@0 279 for j = 1:nc
wolffd@0 280 ee = ee+sum(dk(:,:,j).^2);
wolffd@0 281 end
wolffd@0 282 ee = sqrt(ee/nl/nc);
wolffd@0 283 else
wolffd@0 284 ee = sqrt(sum(ac.sqrsum.^2)/ac.samples);
wolffd@0 285 end
wolffd@0 286 dk = dk./repmat(ee,[nl,1,nc]);
wolffd@0 287 end
wolffd@0 288 if isfield(para,'trim') && not(isequal(para.trim,0)) ...
wolffd@0 289 && not(strcmpi(para.trim,'NoTrim'))
wolffd@0 290 if not(para.trimthreshold)
wolffd@0 291 para.trimthreshold = 0.06;
wolffd@0 292 end
wolffd@0 293 trimframe = 100;
wolffd@0 294 trimhop = 10;
wolffd@0 295 nframes = floor((length(tk)-trimframe)/trimhop)+1;
wolffd@0 296 rms = zeros(1,nframes);
wolffd@0 297 for j = 1:nframes
wolffd@0 298 st = floor((j-1)*trimhop)+1;
wolffd@0 299 for z = 1:size(dk,3)
wolffd@0 300 rms(1,j,z) = norm(dk(st:st+trimframe-1,1,z))/sqrt(trimframe);
wolffd@0 301 end
wolffd@0 302 end
wolffd@0 303 rms = (rms-repmat(min(rms),[1,size(rms,2),1]))...
wolffd@0 304 ./repmat(max(rms)-min(rms),[1,size(rms,2),1]);
wolffd@0 305 nosil = find(rms>para.trimthreshold);
wolffd@0 306 if strcmpi(para.trim,'Trim') || strcmpi(para.trim,'TrimStart') ...
wolffd@0 307 || strcmpi(para.trim,'TrimBegin')
wolffd@0 308 nosil1 = min(nosil);
wolffd@0 309 if nosil1 > 1
wolffd@0 310 nosil1 = nosil1-1;
wolffd@0 311 end
wolffd@0 312 n1 = floor((nosil1-1)*trimhop)+1;
wolffd@0 313 else
wolffd@0 314 n1 = 1;
wolffd@0 315 end
wolffd@0 316 if strcmpi(para.trim,'Trim') || strcmpi(para.trim,'TrimEnd')
wolffd@0 317 nosil2 = max(nosil);
wolffd@0 318 if nosil2 < length(rms)
wolffd@0 319 nosil2 = nosil2+1;
wolffd@0 320 end
wolffd@0 321 n2 = floor((nosil2-1)*trimhop)+1;
wolffd@0 322 else
wolffd@0 323 n2 = length(tk);
wolffd@0 324 end
wolffd@0 325 wh = ones(n2-n1+1,1);
wolffd@0 326 dt = round(.02*f{h});
wolffd@0 327 ha = hann(dt*2);
wolffd@0 328 wh(1:dt) = ha(1:dt);
wolffd@0 329 wh(end-dt+1:end) = ha(dt+1:end);
wolffd@0 330 tk = tk(n1:n2);
wolffd@0 331 dk = dk(n1:n2,1,:);%.*repmat(wh,[1 1 size(dk,3)]);
wolffd@0 332 end
wolffd@0 333 if isfield(para,'sampling') && para.sampling
wolffd@0 334 if and(f{k}, not(f{k} == para.sampling))
wolffd@0 335 for j = 1:size(dk,3)
wolffd@0 336 rk(:,:,j) = resample(dk(:,:,j),para.sampling,f{k});
wolffd@0 337 end
wolffd@0 338 dk = rk;
wolffd@0 339 tk = repmat((0:size(dk,1)-1)',[1 1 size(tk,3)])...
wolffd@0 340 /para.sampling + tk(1,:,:);
wolffd@0 341 end
wolffd@0 342 f{k} = para.sampling;
wolffd@0 343 end
wolffd@0 344 d{h}{k} = dk;
wolffd@0 345 t{h}{k} = tk;
wolffd@0 346 %if isfield(para,'reverse') && para.reverse
wolffd@0 347 % d{h}{k} = flipdim(d{h}{k},1);
wolffd@0 348 %end
wolffd@0 349 end
wolffd@0 350 end
wolffd@0 351 a = set(a,'Data',d,'Time',t,'Sampling',f,'Clusters',cl);
wolffd@0 352 if isfield(para,'label')
wolffd@0 353 if isnumeric(para.label)
wolffd@0 354 n = get(a,'Name');
wolffd@0 355 l = cell(1,length(d));
wolffd@0 356 for k = 1:length(d)
wolffd@0 357 if para.label
wolffd@0 358 l{k} = n{k}(para.label);
wolffd@0 359 else
wolffd@0 360 l{k} = n{k};
wolffd@0 361 end
wolffd@0 362 end
wolffd@0 363 a = set(a,'Label',l);
wolffd@0 364 elseif iscell(para.label)
wolffd@0 365 idx = mod(get(a,'Index'),length(para.label));
wolffd@0 366 if not(idx)
wolffd@0 367 idx = length(para.label);
wolffd@0 368 end
wolffd@0 369 a = set(a,'Label',para.label{idx});
wolffd@0 370 elseif ischar(para.label)
wolffd@0 371 l = cell(1,length(d));
wolffd@0 372 for k = 1:length(d)
wolffd@0 373 l{k} = para.label;
wolffd@0 374 end
wolffd@0 375 a = set(a,'Label',l);
wolffd@0 376 end
wolffd@0 377 end
wolffd@0 378
wolffd@0 379
wolffd@0 380 function [new orig] = beforechunk(orig,option,missing)
wolffd@0 381 option.normal = 0;
wolffd@0 382 a = miraudio(orig,option);
wolffd@0 383 d = get(a,'Data');
wolffd@0 384 old = get(orig,'AcrossChunks');
wolffd@0 385 if isempty(old)
wolffd@0 386 old.sqrsum = 0;
wolffd@0 387 old.samples = 0;
wolffd@0 388 end
wolffd@0 389 new = mircompute(@crossum,d);
wolffd@0 390 new = new{1}{1};
wolffd@0 391 new.sqrsum = old.sqrsum + new.sqrsum;
wolffd@0 392 new.samples = old.samples + new.samples;
wolffd@0 393
wolffd@0 394
wolffd@0 395 function s = crossum(d)
wolffd@0 396 s.sqrsum = sum(d.^2);
wolffd@0 397 s.samples = length(d);
wolffd@0 398
wolffd@0 399
wolffd@0 400 function [y orig] = eachchunk(orig,option,missing)
wolffd@0 401 y = miraudio(orig,option);
wolffd@0 402
wolffd@0 403
wolffd@0 404 function y = combinechunk(old,new)
wolffd@0 405 do = get(old,'Data');
wolffd@0 406 to = get(old,'Time');
wolffd@0 407 dn = get(new,'Data');
wolffd@0 408 tn = get(new,'Time');
wolffd@0 409 y = set(old,'Data',{{[do{1}{1};dn{1}{1}]}},...
wolffd@0 410 'Time',{{[to{1}{1};tn{1}{1}]}});