Mercurial > hg > qm-vamp-plugins

annotate plugins/BarBeatTrack.cpp @ 95:e51152b7ee06

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
* Fixes to problems shown by vamp-plugin-tester
author Chris Cannam <c.cannam@qmul.ac.uk>
date Mon, 23 Mar 2009 16:29:12 +0000
parents 385bec9df059
children 7e5fc22c9ab1
rev   line source
c@89 1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
c@89 2
c@89 3 /*
c@89 4 QM Vamp Plugin Set
c@89 5
c@89 6 Centre for Digital Music, Queen Mary, University of London.
c@89 7 All rights reserved.
c@89 8 */
c@89 9
c@89 10 #include "BarBeatTrack.h"
c@89 11
c@89 12 #include <dsp/onsets/DetectionFunction.h>
c@89 13 #include <dsp/onsets/PeakPicking.h>
c@89 14 #include <dsp/tempotracking/TempoTrackV2.h>
c@89 15 #include <dsp/tempotracking/DownBeat.h>
c@89 16 #include <maths/MathUtilities.h>
c@89 17
c@89 18 using std::string;
c@89 19 using std::vector;
c@89 20 using std::cerr;
c@89 21 using std::endl;
c@89 22
c@89 23 float BarBeatTracker::m_stepSecs = 0.01161; // 512 samples at 44100
c@89 24
c@89 25 class BarBeatTrackerData
c@89 26 {
c@89 27 public:
c@89 28 BarBeatTrackerData(float rate, const DFConfig &config) : dfConfig(config) {
c@89 29 df = new DetectionFunction(config);
c@89 30 // decimation factor aims at resampling to c. 3KHz; must be power of 2
c@89 31 int factor = MathUtilities::nextPowerOfTwo(rate / 3000);
c@95 32 // std::cerr << "BarBeatTrackerData: factor = " << factor << std::endl;
c@89 33 downBeat = new DownBeat(rate, factor, config.stepSize);
c@89 34 }
c@89 35 ~BarBeatTrackerData() {
c@89 36 delete df;
c@89 37 delete downBeat;
c@89 38 }
c@89 39 void reset() {
c@89 40 delete df;
c@89 41 df = new DetectionFunction(dfConfig);
c@89 42 dfOutput.clear();
c@89 43 downBeat->resetAudioBuffer();
c@89 44 origin = Vamp::RealTime::zeroTime;
c@89 45 }
c@89 46
c@89 47 DFConfig dfConfig;
c@89 48 DetectionFunction *df;
c@89 49 DownBeat *downBeat;
c@89 50 vector<double> dfOutput;
c@89 51 Vamp::RealTime origin;
c@89 52 };
c@89 53
c@89 54
c@89 55 BarBeatTracker::BarBeatTracker(float inputSampleRate) :
c@89 56 Vamp::Plugin(inputSampleRate),
c@89 57 m_d(0),
c@89 58 m_bpb(4)
c@89 59 {
c@89 60 }
c@89 61
c@89 62 BarBeatTracker::~BarBeatTracker()
c@89 63 {
c@89 64 delete m_d;
c@89 65 }
c@89 66
c@89 67 string
c@89 68 BarBeatTracker::getIdentifier() const
c@89 69 {
c@89 70 return "qm-barbeattracker";
c@89 71 }
c@89 72
c@89 73 string
c@89 74 BarBeatTracker::getName() const
c@89 75 {
c@89 76 return "Bar and Beat Tracker";
c@89 77 }
c@89 78
c@89 79 string
c@89 80 BarBeatTracker::getDescription() const
c@89 81 {
c@89 82 return "Estimate bar and beat locations";
c@89 83 }
c@89 84
c@89 85 string
c@89 86 BarBeatTracker::getMaker() const
c@89 87 {
c@89 88 return "Queen Mary, University of London";
c@89 89 }
c@89 90
c@89 91 int
c@89 92 BarBeatTracker::getPluginVersion() const
c@89 93 {
c@89 94 return 1;
c@89 95 }
c@89 96
c@89 97 string
c@89 98 BarBeatTracker::getCopyright() const
c@89 99 {
c@89 100 return "Plugin by Matthew Davies, Christian Landone and Chris Cannam. Copyright (c) 2006-2009 QMUL - All Rights Reserved";
c@89 101 }
c@89 102
c@89 103 BarBeatTracker::ParameterList
c@89 104 BarBeatTracker::getParameterDescriptors() const
c@89 105 {
c@89 106 ParameterList list;
c@89 107
c@89 108 ParameterDescriptor desc;
c@89 109
c@89 110 desc.identifier = "bpb";
c@89 111 desc.name = "Beats per Bar";
c@89 112 desc.description = "The number of beats in each bar";
c@89 113 desc.minValue = 2;
c@89 114 desc.maxValue = 16;
c@89 115 desc.defaultValue = 4;
c@89 116 desc.isQuantized = true;
c@89 117 desc.quantizeStep = 1;
c@89 118 list.push_back(desc);
c@89 119
c@89 120 return list;
c@89 121 }
c@89 122
c@89 123 float
c@89 124 BarBeatTracker::getParameter(std::string name) const
c@89 125 {
c@89 126 if (name == "bpb") return m_bpb;
c@89 127 return 0.0;
c@89 128 }
c@89 129
c@89 130 void
c@89 131 BarBeatTracker::setParameter(std::string name, float value)
c@89 132 {
c@89 133 if (name == "bpb") m_bpb = lrintf(value);
c@89 134 }
c@89 135
c@89 136 bool
c@89 137 BarBeatTracker::initialise(size_t channels, size_t stepSize, size_t blockSize)
c@89 138 {
c@89 139 if (m_d) {
c@89 140 delete m_d;
c@89 141 m_d = 0;
c@89 142 }
c@89 143
c@89 144 if (channels < getMinChannelCount() ||
c@89 145 channels > getMaxChannelCount()) {
c@89 146 std::cerr << "BarBeatTracker::initialise: Unsupported channel count: "
c@89 147 << channels << std::endl;
c@89 148 return false;
c@89 149 }
c@89 150
c@89 151 if (stepSize != getPreferredStepSize()) {
c@89 152 std::cerr << "ERROR: BarBeatTracker::initialise: Unsupported step size for this sample rate: "
c@89 153 << stepSize << " (wanted " << (getPreferredStepSize()) << ")" << std::endl;
c@89 154 return false;
c@89 155 }
c@89 156
c@89 157 if (blockSize != getPreferredBlockSize()) {
c@89 158 std::cerr << "WARNING: BarBeatTracker::initialise: Sub-optimal block size for this sample rate: "
c@89 159 << blockSize << " (wanted " << getPreferredBlockSize() << ")" << std::endl;
c@89 160 // return false;
c@89 161 }
c@89 162
c@89 163 DFConfig dfConfig;
c@89 164 dfConfig.DFType = DF_COMPLEXSD;
c@89 165 dfConfig.stepSize = stepSize;
c@89 166 dfConfig.frameLength = blockSize;
c@89 167 dfConfig.dbRise = 3;
c@89 168 dfConfig.adaptiveWhitening = false;
c@89 169 dfConfig.whiteningRelaxCoeff = -1;
c@89 170 dfConfig.whiteningFloor = -1;
c@89 171
c@89 172 m_d = new BarBeatTrackerData(m_inputSampleRate, dfConfig);
c@89 173 m_d->downBeat->setBeatsPerBar(m_bpb);
c@89 174 return true;
c@89 175 }
c@89 176
c@89 177 void
c@89 178 BarBeatTracker::reset()
c@89 179 {
c@89 180 if (m_d) m_d->reset();
c@89 181 }
c@89 182
c@89 183 size_t
c@89 184 BarBeatTracker::getPreferredStepSize() const
c@89 185 {
c@89 186 size_t step = size_t(m_inputSampleRate * m_stepSecs + 0.0001);
c@95 187 if (step < 1) step = 1;
c@89 188 // std::cerr << "BarBeatTracker::getPreferredStepSize: input sample rate is " << m_inputSampleRate << ", step size is " << step << std::endl;
c@89 189 return step;
c@89 190 }
c@89 191
c@89 192 size_t
c@89 193 BarBeatTracker::getPreferredBlockSize() const
c@89 194 {
c@89 195 size_t theoretical = getPreferredStepSize() * 2;
c@89 196
c@89 197 // I think this is not necessarily going to be a power of two, and
c@89 198 // the host might have a problem with that, but I'm not sure we
c@89 199 // can do much about it here
c@89 200 return theoretical;
c@89 201 }
c@89 202
c@89 203 BarBeatTracker::OutputList
c@89 204 BarBeatTracker::getOutputDescriptors() const
c@89 205 {
c@89 206 OutputList list;
c@89 207
c@89 208 OutputDescriptor beat;
c@89 209 beat.identifier = "beats";
c@89 210 beat.name = "Beats";
c@89 211 beat.description = "Beat locations labelled with metrical position";
c@89 212 beat.unit = "";
c@89 213 beat.hasFixedBinCount = true;
c@89 214 beat.binCount = 0;
c@89 215 beat.sampleType = OutputDescriptor::VariableSampleRate;
c@89 216 beat.sampleRate = 1.0 / m_stepSecs;
c@89 217
c@89 218 OutputDescriptor bars;
c@89 219 bars.identifier = "bars";
c@89 220 bars.name = "Bars";
c@89 221 bars.description = "Bar locations";
c@89 222 bars.unit = "";
c@89 223 bars.hasFixedBinCount = true;
c@89 224 bars.binCount = 0;
c@89 225 bars.sampleType = OutputDescriptor::VariableSampleRate;
c@89 226 bars.sampleRate = 1.0 / m_stepSecs;
c@89 227
c@89 228 OutputDescriptor beatcounts;
c@89 229 beatcounts.identifier = "beatcounts";
c@89 230 beatcounts.name = "Beat Count";
c@89 231 beatcounts.description = "Beat counter function";
c@89 232 beatcounts.unit = "";
c@89 233 beatcounts.hasFixedBinCount = true;
c@89 234 beatcounts.binCount = 1;
c@89 235 beatcounts.sampleType = OutputDescriptor::VariableSampleRate;
c@89 236 beatcounts.sampleRate = 1.0 / m_stepSecs;
c@89 237
c@90 238 OutputDescriptor beatsd;
c@90 239 beatsd.identifier = "beatsd";
c@90 240 beatsd.name = "Beat Spectral Difference";
c@90 241 beatsd.description = "Beat spectral difference function used for bar-line detection";
c@90 242 beatsd.unit = "";
c@90 243 beatsd.hasFixedBinCount = true;
c@90 244 beatsd.binCount = 1;
c@90 245 beatsd.sampleType = OutputDescriptor::VariableSampleRate;
c@90 246 beatsd.sampleRate = 1.0 / m_stepSecs;
c@90 247
c@89 248 list.push_back(beat);
c@89 249 list.push_back(bars);
c@89 250 list.push_back(beatcounts);
c@90 251 list.push_back(beatsd);
c@89 252
c@89 253 return list;
c@89 254 }
c@89 255
c@89 256 BarBeatTracker::FeatureSet
c@89 257 BarBeatTracker::process(const float *const *inputBuffers,
c@89 258 Vamp::RealTime timestamp)
c@89 259 {
c@89 260 if (!m_d) {
c@89 261 cerr << "ERROR: BarBeatTracker::process: "
c@89 262 << "BarBeatTracker has not been initialised"
c@89 263 << endl;
c@89 264 return FeatureSet();
c@89 265 }
c@89 266
c@89 267 // We use time domain input, because DownBeat requires it -- so we
c@89 268 // use the time-domain version of DetectionFunction::process which
c@89 269 // does its own FFT. It requires doubles as input, so we need to
c@89 270 // make a temporary copy
c@89 271
c@89 272 // We only support a single input channel
c@89 273
c@89 274 const int fl = m_d->dfConfig.frameLength;
c@89 275 double dfinput[fl];
c@89 276 for (int i = 0; i < fl; ++i) dfinput[i] = inputBuffers[0][i];
c@89 277
c@89 278 double output = m_d->df->process(dfinput);
c@89 279
c@89 280 if (m_d->dfOutput.empty()) m_d->origin = timestamp;
c@89 281
c@93 282 // std::cerr << "df[" << m_d->dfOutput.size() << "] is " << output << std::endl;
c@89 283 m_d->dfOutput.push_back(output);
c@89 284
c@89 285 // Downsample and store the incoming audio block.
c@89 286 // We have an overlap on the incoming audio stream (step size is
c@89 287 // half block size) -- this function is configured to take only a
c@89 288 // step size's worth, so effectively ignoring the overlap. Note
c@89 289 // however that this means we omit the last blocksize - stepsize
c@89 290 // samples completely for the purposes of barline detection
c@89 291 // (hopefully not a problem)
c@89 292 m_d->downBeat->pushAudioBlock(inputBuffers[0]);
c@89 293
c@89 294 return FeatureSet();
c@89 295 }
c@89 296
c@89 297 BarBeatTracker::FeatureSet
c@89 298 BarBeatTracker::getRemainingFeatures()
c@89 299 {
c@89 300 if (!m_d) {
c@89 301 cerr << "ERROR: BarBeatTracker::getRemainingFeatures: "
c@89 302 << "BarBeatTracker has not been initialised"
c@89 303 << endl;
c@89 304 return FeatureSet();
c@89 305 }
c@89 306
c@89 307 return barBeatTrack();
c@89 308 }
c@89 309
c@89 310 BarBeatTracker::FeatureSet
c@89 311 BarBeatTracker::barBeatTrack()
c@89 312 {
c@89 313 vector<double> df;
c@89 314 vector<double> beatPeriod;
c@89 315 vector<double> tempi;
c@89 316
c@89 317 for (size_t i = 2; i < m_d->dfOutput.size(); ++i) { // discard first two elts
c@89 318 df.push_back(m_d->dfOutput[i]);
c@89 319 beatPeriod.push_back(0.0);
c@89 320 }
c@89 321 if (df.empty()) return FeatureSet();
c@89 322
c@89 323 TempoTrackV2 tt(m_inputSampleRate, m_d->dfConfig.stepSize);
c@89 324 tt.calculateBeatPeriod(df, beatPeriod, tempi);
c@89 325
c@89 326 vector<double> beats;
c@89 327 tt.calculateBeats(df, beatPeriod, beats);
c@89 328
c@89 329 vector<int> downbeats;
c@89 330 size_t downLength = 0;
c@89 331 const float *downsampled = m_d->downBeat->getBufferedAudio(downLength);
c@89 332 m_d->downBeat->findDownBeats(downsampled, downLength, beats, downbeats);
c@89 333
c@90 334 vector<double> beatsd;
c@90 335 m_d->downBeat->getBeatSD(beatsd);
c@90 336
c@89 337 // std::cerr << "BarBeatTracker: found downbeats at: ";
c@89 338 // for (int i = 0; i < downbeats.size(); ++i) std::cerr << downbeats[i] << " " << std::endl;
c@89 339
c@89 340 FeatureSet returnFeatures;
c@89 341
c@89 342 char label[20];
c@89 343
c@89 344 int dbi = 0;
c@89 345 int beat = 0;
c@89 346 int bar = 0;
c@89 347
c@89 348 for (size_t i = 0; i < beats.size(); ++i) {
c@89 349
c@89 350 size_t frame = beats[i] * m_d->dfConfig.stepSize;
c@89 351
c@89 352 if (dbi < downbeats.size() && i == downbeats[dbi]) {
c@89 353 beat = 0;
c@89 354 ++bar;
c@89 355 ++dbi;
c@89 356 } else {
c@89 357 ++beat;
c@89 358 }
c@89 359
c@89 360 // outputs are:
c@89 361 //
c@89 362 // 0 -> beats
c@89 363 // 1 -> bars
c@89 364 // 2 -> beat counter function
c@89 365
c@89 366 Feature feature;
c@89 367 feature.hasTimestamp = true;
c@89 368 feature.timestamp = m_d->origin + Vamp::RealTime::frame2RealTime
c@89 369 (frame, lrintf(m_inputSampleRate));
c@89 370
c@89 371 sprintf(label, "%d", beat + 1);
c@89 372 feature.label = label;
c@89 373 returnFeatures[0].push_back(feature); // labelled beats
c@89 374
c@89 375 feature.values.push_back(beat + 1);
c@89 376 returnFeatures[2].push_back(feature); // beat function
c@89 377
c@90 378 if (i > 0 && i <= beatsd.size()) {
c@90 379 feature.values.clear();
c@90 380 feature.values.push_back(beatsd[i-1]);
c@90 381 feature.label = "";
c@90 382 returnFeatures[3].push_back(feature); // beat spectral difference
c@90 383 }
c@90 384
c@89 385 if (beat == 0) {
c@89 386 feature.values.clear();
c@89 387 sprintf(label, "%d", bar);
c@89 388 feature.label = label;
c@89 389 returnFeatures[1].push_back(feature); // bars
c@89 390 }
c@89 391 }
c@89 392
c@89 393 return returnFeatures;
c@89 394 }
c@89 395