Vamp feature extraction plugins using Paul Brossier's Aubio library.
Centre for Digital Music, Queen Mary, University of London.
- This file copyright 2006 Chris Cannam.
+ This file copyright 2006-2008 Chris Cannam and QMUL.
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as
#include <math.h>
#include "Notes.h"
+#include <algorithm>
+
using std::string;
using std::vector;
using std::cerr;
Notes::Notes(float inputSampleRate) :
Plugin(inputSampleRate),
m_ibuf(0),
- m_fftgrain(0),
m_onset(0),
- m_pv(0),
- m_peakpick(0),
+ m_pitch(0),
m_onsetdet(0),
- m_onsettype(aubio_onset_complex),
+ m_onsettype(OnsetComplex),
m_pitchdet(0),
- m_pitchtype(aubio_pitch_yinfft),
- m_pitchmode(aubio_pitchm_freq),
+ m_pitchtype(PitchYinFFT),
m_threshold(0.3),
- m_silence(-90),
- m_median(6)
+ m_silence(-70),
+ m_minioi(4),
+ m_median(6),
+ m_minpitch(32),
+ m_maxpitch(95),
+ m_wrapRange(false),
+ m_avoidLeaps(false),
+ m_prevPitch(-1)
{
}
Notes::~Notes()
{
- if (m_onsetdet) aubio_onsetdetection_free(m_onsetdet);
- if (m_pitchdet) del_aubio_pitchdetection(m_pitchdet);
+ if (m_onsetdet) del_aubio_onset(m_onsetdet);
+ if (m_pitchdet) del_aubio_pitch(m_pitchdet);
if (m_ibuf) del_fvec(m_ibuf);
if (m_onset) del_fvec(m_onset);
- if (m_fftgrain) del_cvec(m_fftgrain);
- if (m_pv) del_aubio_pvoc(m_pv);
- if (m_peakpick) del_aubio_peakpicker(m_peakpick);
+ if (m_pitch) del_fvec(m_pitch);
}
string
-Notes::getName() const
+Notes::getIdentifier() const
{
return "aubionotes";
}
string
-Notes::getDescription() const
+Notes::getName() const
{
return "Aubio Note Tracker";
}
string
+Notes::getDescription() const
+{
+ return "Estimate note onset positions, pitches and durations";
+}
+
+string
Notes::getMaker() const
{
return "Paul Brossier (plugin by Chris Cannam)";
int
Notes::getPluginVersion() const
{
- return 1;
+ return 4;
}
string
bool
Notes::initialise(size_t channels, size_t stepSize, size_t blockSize)
{
- m_channelCount = channels;
+ if (channels != 1) {
+ std::cerr << "Notes::initialise: channels must be 1" << std::endl;
+ return false;
+ }
+
m_stepSize = stepSize;
m_blockSize = blockSize;
- size_t processingBlockSize;
- if (m_onsettype == aubio_onset_energy ||
- m_onsettype == aubio_onset_hfc) {
- processingBlockSize = stepSize * 2;
- } else {
- processingBlockSize = stepSize * 4;
- }
+ m_ibuf = new_fvec(stepSize);
+ m_onset = new_fvec(1);
+ m_pitch = new_fvec(1);
- m_ibuf = new_fvec(stepSize, channels);
- m_onset = new_fvec(1, channels);
- m_fftgrain = new_cvec(processingBlockSize, channels);
- m_pv = new_aubio_pvoc(processingBlockSize, stepSize, channels);
- m_peakpick = new_aubio_peakpicker(m_threshold);
+ m_onsetdet = new_aubio_onset
+ (const_cast<char *>(getAubioNameForOnsetType(m_onsettype)),
+ blockSize,
+ stepSize,
+ lrintf(m_inputSampleRate));
+
+ aubio_onset_set_threshold(m_onsetdet, m_threshold);
+ aubio_onset_set_silence(m_onsetdet, m_silence);
+ aubio_onset_set_minioi(m_onsetdet, m_minioi);
- m_onsetdet = new_aubio_onsetdetection(m_onsettype, processingBlockSize, channels);
+ m_pitchdet = new_aubio_pitch
+ (const_cast<char *>(getAubioNameForPitchType(m_pitchtype)),
+ blockSize,
+ stepSize,
+ lrintf(m_inputSampleRate));
- m_pitchdet = new_aubio_pitchdetection(processingBlockSize * 4,
- stepSize,
- channels,
- lrintf(m_inputSampleRate),
- m_pitchtype,
- m_pitchmode);
+ aubio_pitch_set_unit(m_pitchdet, const_cast<char *>("freq"));
m_count = 0;
m_delay = Vamp::RealTime::frame2RealTime((4 + m_median) * m_stepSize,
lrintf(m_inputSampleRate));
m_currentOnset = Vamp::RealTime::zeroTime;
m_haveCurrent = false;
+ m_prevPitch = -1;
return true;
}
ParameterList list;
ParameterDescriptor desc;
- desc.name = "onsettype";
- desc.description = "Onset Detection Function Type";
+ desc.identifier = "onsettype";
+ desc.name = "Onset Detection Function Type";
desc.minValue = 0;
- desc.maxValue = 6;
- desc.defaultValue = (int)aubio_onset_complex;
+ desc.maxValue = 7;
+ desc.defaultValue = (int)OnsetComplex;
desc.isQuantized = true;
desc.quantizeStep = 1;
desc.valueNames.push_back("Energy Based");
desc.valueNames.push_back("Phase Deviation");
desc.valueNames.push_back("Kullback-Liebler");
desc.valueNames.push_back("Modified Kullback-Liebler");
+ desc.valueNames.push_back("Spectral Flux");
list.push_back(desc);
desc = ParameterDescriptor();
- desc.name = "pitchtype";
- desc.description = "Pitch Detection Function Type";
+ desc.identifier = "pitchtype";
+ desc.name = "Pitch Detection Function Type";
desc.minValue = 0;
desc.maxValue = 4;
- desc.defaultValue = (int)aubio_pitch_yinfft;
+ desc.defaultValue = (int)PitchYinFFT;
desc.isQuantized = true;
desc.quantizeStep = 1;
desc.valueNames.push_back("YIN Frequency Estimator");
list.push_back(desc);
desc = ParameterDescriptor();
- desc.name = "peakpickthreshold";
- desc.description = "Peak Picker Threshold";
+ desc.identifier = "minpitch";
+ desc.name = "Minimum Pitch";
+ desc.minValue = 0;
+ desc.maxValue = 127;
+ desc.defaultValue = 32;
+ desc.unit = "MIDI units";
+ desc.isQuantized = true;
+ desc.quantizeStep = 1;
+ list.push_back(desc);
+
+ desc = ParameterDescriptor();
+ desc.identifier = "maxpitch";
+ desc.name = "Maximum Pitch";
+ desc.minValue = 0;
+ desc.maxValue = 127;
+ desc.defaultValue = 95;
+ desc.unit = "MIDI units";
+ desc.isQuantized = true;
+ desc.quantizeStep = 1;
+ list.push_back(desc);
+
+ desc = ParameterDescriptor();
+ desc.identifier = "wraprange";
+ desc.name = "Fold Higher or Lower Notes into Range";
+ desc.minValue = 0;
+ desc.maxValue = 1;
+ desc.defaultValue = 0;
+ desc.isQuantized = true;
+ desc.quantizeStep = 1;
+ list.push_back(desc);
+
+ desc = ParameterDescriptor();
+ desc.identifier = "avoidleaps";
+ desc.name = "Avoid Multi-Octave Jumps";
+ desc.minValue = 0;
+ desc.maxValue = 1;
+ desc.defaultValue = 0;
+ desc.isQuantized = true;
+ desc.quantizeStep = 1;
+ list.push_back(desc);
+
+ desc = ParameterDescriptor();
+ desc.identifier = "peakpickthreshold";
+ desc.name = "Peak Picker Threshold";
desc.minValue = 0;
desc.maxValue = 1;
desc.defaultValue = 0.3;
list.push_back(desc);
desc = ParameterDescriptor();
- desc.name = "silencethreshold";
- desc.description = "Silence Threshold";
+ desc.identifier = "silencethreshold";
+ desc.name = "Silence Threshold";
desc.minValue = -120;
desc.maxValue = 0;
- desc.defaultValue = -90;
+ desc.defaultValue = -70;
desc.unit = "dB";
desc.isQuantized = false;
list.push_back(desc);
+ desc = ParameterDescriptor();
+ desc.identifier = "minioi";
+ desc.name = "Minimum Inter-Onset Interval";
+ desc.minValue = 0;
+ desc.maxValue = 40;
+ desc.defaultValue = 4;
+ desc.unit = "ms";
+ desc.isQuantized = true;
+ desc.quantizeStep = 1;
+ list.push_back(desc);
+
return list;
}
return m_threshold;
} else if (param == "silencethreshold") {
return m_silence;
+ } else if (param == "minpitch") {
+ return m_minpitch;
+ } else if (param == "maxpitch") {
+ return m_maxpitch;
+ } else if (param == "wraprange") {
+ return m_wrapRange ? 1.0 : 0.0;
+ } else if (param == "avoidleaps") {
+ return m_avoidLeaps ? 1.0 : 0.0;
+ } else if (param == "minioi") {
+ return m_minioi;
} else {
return 0.0;
}
{
if (param == "onsettype") {
switch (lrintf(value)) {
- case 0: m_onsettype = aubio_onset_energy; break;
- case 1: m_onsettype = aubio_onset_specdiff; break;
- case 2: m_onsettype = aubio_onset_hfc; break;
- case 3: m_onsettype = aubio_onset_complex; break;
- case 4: m_onsettype = aubio_onset_phase; break;
- case 5: m_onsettype = aubio_onset_kl; break;
- case 6: m_onsettype = aubio_onset_mkl; break;
+ case 0: m_onsettype = OnsetEnergy; break;
+ case 1: m_onsettype = OnsetSpecDiff; break;
+ case 2: m_onsettype = OnsetHFC; break;
+ case 3: m_onsettype = OnsetComplex; break;
+ case 4: m_onsettype = OnsetPhase; break;
+ case 5: m_onsettype = OnsetKL; break;
+ case 6: m_onsettype = OnsetMKL; break;
+ case 7: m_onsettype = OnsetSpecFlux; break;
}
} else if (param == "pitchtype") {
switch (lrintf(value)) {
- case 0: m_pitchtype = aubio_pitch_yin; break;
- case 1: m_pitchtype = aubio_pitch_mcomb; break;
- case 2: m_pitchtype = aubio_pitch_schmitt; break;
- case 3: m_pitchtype = aubio_pitch_fcomb; break;
- case 4: m_pitchtype = aubio_pitch_yinfft; break;
+ case 0: m_pitchtype = PitchYin; break;
+ case 1: m_pitchtype = PitchMComb; break;
+ case 2: m_pitchtype = PitchSchmitt; break;
+ case 3: m_pitchtype = PitchFComb; break;
+ case 4: m_pitchtype = PitchYinFFT; break;
}
} else if (param == "peakpickthreshold") {
m_threshold = value;
} else if (param == "silencethreshold") {
m_silence = value;
+ } else if (param == "minpitch") {
+ m_minpitch = lrintf(value);
+ } else if (param == "maxpitch") {
+ m_maxpitch = lrintf(value);
+ } else if (param == "wraprange") {
+ m_wrapRange = (value > 0.5);
+ } else if (param == "avoidleaps") {
+ m_avoidLeaps = (value > 0.5);
+ } else if (param == "minioi") {
+ m_minioi = value;
}
}
OutputList list;
OutputDescriptor d;
- d.name = "notes";
+ d.identifier = "notes";
+ d.name = "Notes";
d.unit = "Hz";
- d.description = "Notes";
d.hasFixedBinCount = true;
+
d.binCount = 2;
d.binNames.push_back("Frequency");
- d.binNames.push_back("Duration");
d.binNames.push_back("Velocity");
+ d.hasDuration = true;
+
d.hasKnownExtents = false;
d.isQuantized = false;
d.sampleType = OutputDescriptor::VariableSampleRate;
}
Notes::FeatureSet
-Notes::process(float **inputBuffers, Vamp::RealTime timestamp)
+Notes::process(const float *const *inputBuffers, Vamp::RealTime timestamp)
{
for (size_t i = 0; i < m_stepSize; ++i) {
- for (size_t j = 0; j < m_channelCount; ++j) {
- fvec_write_sample(m_ibuf, inputBuffers[j][i], j, i);
- }
+ fvec_write_sample(m_ibuf, inputBuffers[0][i], i);
}
- aubio_pvoc_do(m_pv, m_ibuf, m_fftgrain);
- aubio_onsetdetection(m_onsetdet, m_fftgrain, m_onset);
-
- bool isonset = aubio_peakpick_pimrt(m_onset, m_peakpick);
+ aubio_onset_do(m_onsetdet, m_ibuf, m_onset);
+ aubio_pitch_do(m_pitchdet, m_ibuf, m_pitch);
- float frequency = aubio_pitchdetection(m_pitchdet, m_ibuf);
+ bool isonset = m_onset->data[0];
+ float frequency = m_pitch->data[0];
m_notebuf.push_back(frequency);
if (m_notebuf.size() > m_median) m_notebuf.pop_front();
float median = toSort[toSort.size()/2];
if (median < 45.0) return;
+ float freq = median;
+ int midiPitch = (int)floor(aubio_freqtomidi(freq) + 0.5);
+
+ if (m_avoidLeaps) {
+ if (m_prevPitch >= 0) {
+ while (midiPitch < m_prevPitch - 12) {
+ midiPitch += 12;
+ freq *= 2;
+ }
+ while (midiPitch > m_prevPitch + 12) {
+ midiPitch -= 12;
+ freq /= 2;
+ }
+ }
+ }
+
+ while (midiPitch < m_minpitch) {
+ if (!m_wrapRange) return;
+ midiPitch += 12;
+ freq *= 2;
+ }
+
+ while (midiPitch > m_maxpitch) {
+ if (!m_wrapRange) return;
+ midiPitch -= 12;
+ freq /= 2;
+ }
+
+ m_prevPitch = midiPitch;
+
Feature feature;
feature.hasTimestamp = true;
if (m_currentOnset < m_delay) m_currentOnset = m_delay;
feature.timestamp = m_currentOnset - m_delay;
- feature.values.push_back(median);
-// feature.values.push_back(FLOOR(aubio_freqtomidi(median) + 0.5));
+ feature.values.push_back(freq);
+
feature.values.push_back
- (Vamp::RealTime::realTime2Frame(offTime, lrintf(m_inputSampleRate)) -
- Vamp::RealTime::realTime2Frame(m_currentOnset, lrintf(m_inputSampleRate)));
+ (Vamp::RealTime::realTime2Frame
+ (offTime, lrintf(m_inputSampleRate)) -
+ Vamp::RealTime::realTime2Frame
+ (m_currentOnset, lrintf(m_inputSampleRate)));
+ feature.hasDuration = false;
+
feature.values.push_back(m_currentLevel);
fs[0].push_back(feature);
}
projects / vamp-aubio-plugins.git / blobdiff