| 1 | """Copyright (C) 2004 Paul Brossier <piem@altern.org> |
|---|
| 2 | print aubio.__LICENSE__ for the terms of use |
|---|
| 3 | """ |
|---|
| 4 | |
|---|
| 5 | __LICENSE__ = """\ |
|---|
| 6 | Copyright (C) 2004 Paul Brossier <piem@altern.org> |
|---|
| 7 | |
|---|
| 8 | This program is free software; you can redistribute it and/or modify |
|---|
| 9 | it under the terms of the GNU General Public License as published by |
|---|
| 10 | the Free Software Foundation; either version 2 of the License, or |
|---|
| 11 | (at your option) any later version. |
|---|
| 12 | |
|---|
| 13 | This program is distributed in the hope that it will be useful, |
|---|
| 14 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
|---|
| 15 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
|---|
| 16 | GNU General Public License for more details. |
|---|
| 17 | |
|---|
| 18 | You should have received a copy of the GNU General Public License |
|---|
| 19 | along with this program; if not, write to the Free Software |
|---|
| 20 | Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
|---|
| 21 | """ |
|---|
| 22 | |
|---|
| 23 | |
|---|
| 24 | __notesheight = 0.25 |
|---|
| 25 | |
|---|
| 26 | |
|---|
| 27 | def audio_to_array(filename): |
|---|
| 28 | import aubio.aubioclass |
|---|
| 29 | import numarray |
|---|
| 30 | hopsize = 2048 |
|---|
| 31 | filei = aubio.aubioclass.sndfile(filename) |
|---|
| 32 | framestep = 1/(filei.samplerate()+0.) |
|---|
| 33 | channels = filei.channels() |
|---|
| 34 | myvec = aubio.aubioclass.fvec(hopsize,channels) |
|---|
| 35 | data = [] |
|---|
| 36 | readsize = hopsize |
|---|
| 37 | while (readsize==hopsize): |
|---|
| 38 | readsize = filei.read(hopsize,myvec) |
|---|
| 39 | #for i in range(channels): |
|---|
| 40 | i = 0 |
|---|
| 41 | curpos = 0 |
|---|
| 42 | while (curpos < readsize): |
|---|
| 43 | data.append(myvec.get(curpos,i)) |
|---|
| 44 | curpos+=1 |
|---|
| 45 | time = numarray.arange(len(data))*framestep |
|---|
| 46 | return time,data |
|---|
| 47 | |
|---|
| 48 | def plot_audio(filenames, fileout=None, start=0, end=None, noaxis=None): |
|---|
| 49 | g = gnuplot_init(fileout) |
|---|
| 50 | d = [] |
|---|
| 51 | todraw = len(filenames) |
|---|
| 52 | xorig = 0. |
|---|
| 53 | xsize = 1./todraw |
|---|
| 54 | g.gnuplot('set multiplot;') |
|---|
| 55 | while (len(filenames)): |
|---|
| 56 | time,data = audio_to_array(filenames.pop(0)) |
|---|
| 57 | d.append(make_audio_plot(time,data)) |
|---|
| 58 | if not noaxis and todraw==1: |
|---|
| 59 | g.xlabel('Time (s)') |
|---|
| 60 | g.ylabel('Amplitude') |
|---|
| 61 | g.gnuplot('set size %f,1.;' % (xsize) ) |
|---|
| 62 | g.gnuplot('set origin %f,0.;' % (xorig) ) |
|---|
| 63 | g.gnuplot('set style data lines; \ |
|---|
| 64 | set yrange [-1.:1.]; \ |
|---|
| 65 | set xrange [0:%f]' % time[-1]) |
|---|
| 66 | g.plot(d.pop(0)) |
|---|
| 67 | xorig += 1./todraw |
|---|
| 68 | g.gnuplot('unset multiplot;') |
|---|
| 69 | |
|---|
| 70 | def audio_to_spec(filename): |
|---|
| 71 | from aubioclass import fvec,cvec,pvoc,sndfile |
|---|
| 72 | from math import log |
|---|
| 73 | bufsize = 256*16 |
|---|
| 74 | hopsize = bufsize/4 # could depend on filelength |
|---|
| 75 | filei = sndfile(filename) |
|---|
| 76 | srate = float(filei.samplerate()) |
|---|
| 77 | framestep = hopsize/srate |
|---|
| 78 | freqstep = srate/bufsize |
|---|
| 79 | channels = filei.channels() |
|---|
| 80 | myvec = fvec(hopsize,channels) |
|---|
| 81 | myfft = cvec(bufsize,channels) |
|---|
| 82 | pv = pvoc(bufsize,hopsize,channels) |
|---|
| 83 | data,time,freq = [],[],[] |
|---|
| 84 | for f in range(bufsize/2): |
|---|
| 85 | freq.append(f*freqstep) |
|---|
| 86 | readsize = hopsize |
|---|
| 87 | frameread = 0 |
|---|
| 88 | while (readsize==hopsize): |
|---|
| 89 | readsize = filei.read(hopsize,myvec) |
|---|
| 90 | pv.do(myvec,myfft) |
|---|
| 91 | frame = [] |
|---|
| 92 | i = 0 #for i in range(channels): |
|---|
| 93 | curpos = 0 |
|---|
| 94 | while (curpos < bufsize/2): |
|---|
| 95 | frame.append(log(myfft.get(curpos,i)**2+0.000001)) |
|---|
| 96 | curpos+=1 |
|---|
| 97 | time.append(frameread*framestep) |
|---|
| 98 | data.append(frame) |
|---|
| 99 | frameread += 1 |
|---|
| 100 | # crop data if unfinished frames |
|---|
| 101 | if len(data[-1]) != len(data[0]): |
|---|
| 102 | data = data[0:-2] |
|---|
| 103 | time = time[0:-2] |
|---|
| 104 | # verify size consistency |
|---|
| 105 | assert len(data) == len(time) |
|---|
| 106 | assert len(data[0]) == len(freq) |
|---|
| 107 | return data,time,freq |
|---|
| 108 | |
|---|
| 109 | def plot_spec(filenames, outplot='',extension='', fileout=None, start=0, end=None, noaxis=None,log=1): |
|---|
| 110 | import Gnuplot |
|---|
| 111 | g = gnuplot_create(outplot,extension) |
|---|
| 112 | todraw = len(filenames) |
|---|
| 113 | xorig = 0. |
|---|
| 114 | xsize = 1./todraw |
|---|
| 115 | data,time,freq = audio_to_spec(filenames.pop(0)) |
|---|
| 116 | |
|---|
| 117 | if not noaxis and todraw==1: |
|---|
| 118 | g.xlabel('Time (s)') |
|---|
| 119 | g.ylabel('Frequency (Hz)') |
|---|
| 120 | g.gnuplot('set pm3d map') |
|---|
| 121 | #g.gnuplot('set palette rgbformulae 30,31,32') |
|---|
| 122 | #g.gnuplot('set palette') |
|---|
| 123 | g.gnuplot('set xrange [0.:%f]' % time[-1]) |
|---|
| 124 | g.gnuplot('set yrange [1.:%f]' % (freq[-1]/1.)) |
|---|
| 125 | if log: |
|---|
| 126 | g.gnuplot('set yrange [10.1:%f]' % (freq[-1]/1.)) |
|---|
| 127 | g.gnuplot('set log y') |
|---|
| 128 | g.splot(Gnuplot.GridData(data,time,freq, binary=1)) |
|---|
| 129 | xorig += 1./todraw |
|---|
| 130 | |
|---|
| 131 | def downsample_audio(time,data,maxpoints=10000): |
|---|
| 132 | """ resample audio data to last only maxpoints """ |
|---|
| 133 | import numarray |
|---|
| 134 | length = len(time) |
|---|
| 135 | downsample = length/maxpoints |
|---|
| 136 | if downsample == 0: downsample = 1 |
|---|
| 137 | x = numarray.array(time).resize(length)[0:-1:downsample] |
|---|
| 138 | y = numarray.array(data).resize(length)[0:-1:downsample] |
|---|
| 139 | return x,y |
|---|
| 140 | |
|---|
| 141 | def make_audio_plot(time,data,maxpoints=10000): |
|---|
| 142 | """ create gnuplot plot from an audio file """ |
|---|
| 143 | import numarray |
|---|
| 144 | import Gnuplot, Gnuplot.funcutils |
|---|
| 145 | length = len(time) |
|---|
| 146 | downsample = length/maxpoints |
|---|
| 147 | if downsample == 0: downsample = 1 |
|---|
| 148 | x = numarray.array(time).resize(length)[0:-1:downsample] |
|---|
| 149 | y = numarray.array(data).resize(length)[0:-1:downsample] |
|---|
| 150 | return Gnuplot.Data(x,y,with='lines') |
|---|
| 151 | |
|---|
| 152 | |
|---|
| 153 | def plot_onsets(filename, onsets, ofunc, samplerate=44100., hopsize=512, outplot=None): |
|---|
| 154 | import Gnuplot, Gnuplot.funcutils |
|---|
| 155 | import aubio.txtfile |
|---|
| 156 | import os.path |
|---|
| 157 | import numarray |
|---|
| 158 | import re |
|---|
| 159 | from aubio.onsetcompare import onset_roc |
|---|
| 160 | |
|---|
| 161 | d,d2 = [],[] |
|---|
| 162 | maxofunc = 0 |
|---|
| 163 | for i in range(len(onsets)): |
|---|
| 164 | if len(onsets[i]) == 0: onsets[i] = [0.]; |
|---|
| 165 | |
|---|
| 166 | # onset detection function |
|---|
| 167 | downtime = (hopsize/samplerate)*numarray.arange(len(ofunc[i])) |
|---|
| 168 | d.append(Gnuplot.Data(downtime,ofunc[i],with='lines')) |
|---|
| 169 | maxofunc = max(max(ofunc[i]), maxofunc) |
|---|
| 170 | |
|---|
| 171 | for i in range(len(onsets)): |
|---|
| 172 | # detected onsets |
|---|
| 173 | x1 = (hopsize/samplerate)*numarray.array(onsets[i]) |
|---|
| 174 | y1 = maxofunc*numarray.ones(len(onsets[i])) |
|---|
| 175 | d.append(Gnuplot.Data(x1,y1,with='impulses')) |
|---|
| 176 | d2.append(Gnuplot.Data(x1,-y1,with='impulses')) |
|---|
| 177 | |
|---|
| 178 | # check if datafile exists truth |
|---|
| 179 | datafile = filename.replace('.wav','.txt') |
|---|
| 180 | if datafile == filename: datafile = "" |
|---|
| 181 | if not os.path.isfile(datafile): |
|---|
| 182 | title = "truth file not found" |
|---|
| 183 | t = Gnuplot.Data(0,0,with='impulses') |
|---|
| 184 | else: |
|---|
| 185 | t_onsets = aubio.txtfile.read_datafile(datafile) |
|---|
| 186 | y2 = maxofunc*numarray.ones(len(t_onsets)) |
|---|
| 187 | x2 = numarray.array(t_onsets).resize(len(t_onsets)) |
|---|
| 188 | d2.append(Gnuplot.Data(x2,y2,with='impulses')) |
|---|
| 189 | |
|---|
| 190 | tol = 0.050 |
|---|
| 191 | |
|---|
| 192 | orig, missed, merged, expc, bad, doubled = \ |
|---|
| 193 | onset_roc(x2,x1,tol) |
|---|
| 194 | title = "GD %2.3f%% FP %2.3f%%" % \ |
|---|
| 195 | ((100*float(orig-missed-merged)/(orig)), |
|---|
| 196 | (100*float(bad+doubled)/(orig))) |
|---|
| 197 | |
|---|
| 198 | # audio data |
|---|
| 199 | time,data = audio_to_array(filename) |
|---|
| 200 | d2.append(make_audio_plot(time,data)) |
|---|
| 201 | |
|---|
| 202 | # prepare the plot |
|---|
| 203 | g = gnuplot_init(outplot) |
|---|
| 204 | |
|---|
| 205 | g('set title \'%s %s\'' % (re.sub('.*/','',filename),title)) |
|---|
| 206 | |
|---|
| 207 | g('set multiplot') |
|---|
| 208 | |
|---|
| 209 | # hack to align left axis |
|---|
| 210 | g('set lmargin 15') |
|---|
| 211 | |
|---|
| 212 | # plot waveform and onsets |
|---|
| 213 | g('set size 1,0.3') |
|---|
| 214 | g('set origin 0,0.7') |
|---|
| 215 | g('set xrange [0:%f]' % max(time)) |
|---|
| 216 | g('set yrange [-1:1]') |
|---|
| 217 | g.ylabel('amplitude') |
|---|
| 218 | g.plot(*d2) |
|---|
| 219 | |
|---|
| 220 | g('unset title') |
|---|
| 221 | |
|---|
| 222 | # plot onset detection function |
|---|
| 223 | g('set size 1,0.7') |
|---|
| 224 | g('set origin 0,0') |
|---|
| 225 | g('set xrange [0:%f]' % (hopsize/samplerate*len(ofunc[0]))) |
|---|
| 226 | g('set yrange [0:%f]' % (maxofunc*1.01)) |
|---|
| 227 | g.xlabel('time') |
|---|
| 228 | g.ylabel('onset detection value') |
|---|
| 229 | g.plot(*d) |
|---|
| 230 | |
|---|
| 231 | g('unset multiplot') |
|---|
| 232 | |
|---|
| 233 | |
|---|
| 234 | def plot_pitch(filename, pitch, samplerate=44100., hopsize=512, outplot=None): |
|---|
| 235 | import aubio.txtfile |
|---|
| 236 | import os.path |
|---|
| 237 | import numarray |
|---|
| 238 | import Gnuplot |
|---|
| 239 | import re |
|---|
| 240 | |
|---|
| 241 | d = [] |
|---|
| 242 | maxpitch = 1000. |
|---|
| 243 | for i in range(len(pitch)): |
|---|
| 244 | #if len(pitch[i]) == 0: pitch[i] = [0.]; |
|---|
| 245 | |
|---|
| 246 | downtime = (hopsize/samplerate)*numarray.arange(len(pitch[i])) |
|---|
| 247 | d.append(Gnuplot.Data(downtime,pitch[i],with='lines', |
|---|
| 248 | title=('%d' % i))) |
|---|
| 249 | maxpitch = max(maxpitch,max(pitch[i][:])*1.1) |
|---|
| 250 | |
|---|
| 251 | # check if ground truth exists |
|---|
| 252 | datafile = filename.replace('.wav','.txt') |
|---|
| 253 | if datafile == filename: datafile = "" |
|---|
| 254 | if not os.path.isfile(datafile): |
|---|
| 255 | title = "truth file not found" |
|---|
| 256 | t = Gnuplot.Data(0,0,with='impulses') |
|---|
| 257 | else: |
|---|
| 258 | title = "truth file plotting not implemented yet" |
|---|
| 259 | values = aubio.txtfile.read_datafile(datafile) |
|---|
| 260 | if (len(datafile[0])) > 1: |
|---|
| 261 | time, pitch = [], [] |
|---|
| 262 | for i in range(len(values)): |
|---|
| 263 | time.append(values[i][0]) |
|---|
| 264 | pitch.append(values[i][1]) |
|---|
| 265 | d.append(Gnuplot.Data(time,pitch,with='lines', |
|---|
| 266 | title='ground truth')) |
|---|
| 267 | |
|---|
| 268 | # audio data |
|---|
| 269 | time,data = audio_to_array(filename) |
|---|
| 270 | f = make_audio_plot(time,data) |
|---|
| 271 | |
|---|
| 272 | g = gnuplot_init(outplot) |
|---|
| 273 | g('set title \'%s %s\'' % (re.sub('.*/','',filename),title)) |
|---|
| 274 | g('set multiplot') |
|---|
| 275 | # hack to align left axis |
|---|
| 276 | g('set lmargin 15') |
|---|
| 277 | # plot waveform and onsets |
|---|
| 278 | g('set size 1,0.3') |
|---|
| 279 | g('set origin 0,0.7') |
|---|
| 280 | g('set xrange [0:%f]' % max(time)) |
|---|
| 281 | g('set yrange [-1:1]') |
|---|
| 282 | g.ylabel('amplitude') |
|---|
| 283 | g.plot(f) |
|---|
| 284 | g('unset title') |
|---|
| 285 | # plot onset detection function |
|---|
| 286 | g('set size 1,0.7') |
|---|
| 287 | g('set origin 0,0') |
|---|
| 288 | g('set xrange [0:%f]' % max(time)) |
|---|
| 289 | g('set yrange [40:%f]' % maxpitch) |
|---|
| 290 | g('set key right top') |
|---|
| 291 | g('set noclip one') |
|---|
| 292 | g.xlabel('time') |
|---|
| 293 | g.ylabel('frequency (Hz)') |
|---|
| 294 | g.plot(*d) |
|---|
| 295 | g('unset multiplot') |
|---|
| 296 | |
|---|
| 297 | def gnuplot_init(outplot,debug=0,persist=1): |
|---|
| 298 | # prepare the plot |
|---|
| 299 | import Gnuplot |
|---|
| 300 | g = Gnuplot.Gnuplot(debug=debug, persist=persist) |
|---|
| 301 | if outplot == 'stdout': |
|---|
| 302 | g("set terminal png fontfile 'p052023l.pfb'") |
|---|
| 303 | #g('set output \'%s\'' % outplot) |
|---|
| 304 | elif outplot: |
|---|
| 305 | extension = outplot.split('.')[-1] |
|---|
| 306 | if extension == 'ps': extension = 'postscript' |
|---|
| 307 | g('set terminal %s' % extension) |
|---|
| 308 | g('set output \'%s\'' % outplot) |
|---|
| 309 | return g |
|---|
| 310 | |
|---|
| 311 | def gnuplot_create(outplot='',extension='',debug=0,persist=1): |
|---|
| 312 | import Gnuplot |
|---|
| 313 | g = Gnuplot.Gnuplot(debug=debug, persist=persist) |
|---|
| 314 | if not extension or not outplot: return g |
|---|
| 315 | if extension == 'ps': ext, extension = '.ps' , 'postscript' |
|---|
| 316 | elif extension == 'png': ext, extension = '.png', 'png' |
|---|
| 317 | elif extension == 'svg': ext, extension = '.svg', 'svg' |
|---|
| 318 | else: exit("ERR: unknown plot extension") |
|---|
| 319 | g('set terminal %s' % extension) |
|---|
| 320 | if outplot != "stdout": |
|---|
| 321 | g('set output \'roc-%s%s\'' % (outplot,ext)) |
|---|
| 322 | return g |
|---|
