cannam@0: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
cannam@0: 
cannam@0: /*
cannam@0:     Vamp feature extraction plugins using Paul Brossier's Aubio library.
cannam@0: 
cannam@0:     Centre for Digital Music, Queen Mary, University of London.
cannam@20:     This file copyright 2006-2008 Chris Cannam and QMUL.
cannam@0:     
piem@112:     This file is part of vamp-aubio-plugins.
piem@112: 
piem@112:     vamp-aubio is free software: you can redistribute it and/or modify
piem@112:     it under the terms of the GNU General Public License as published by
piem@112:     the Free Software Foundation, either version 3 of the License, or
piem@112:     (at your option) any later version.
piem@112: 
piem@112:     vamp-aubio is distributed in the hope that it will be useful,
piem@112:     but WITHOUT ANY WARRANTY; without even the implied warranty of
piem@112:     MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
piem@112:     GNU General Public License for more details.
piem@112: 
piem@112:     You should have received a copy of the GNU General Public License
piem@112:     along with aubio.  If not, see <http://www.gnu.org/licenses/>.
cannam@0: 
cannam@0: */
cannam@0: 
piem@2: #include <math.h>
cannam@0: #include "Notes.h"
cannam@0: 
cannam@28: #include <algorithm>
cannam@28: 
cannam@0: using std::string;
cannam@0: using std::vector;
cannam@0: using std::cerr;
cannam@0: using std::endl;
cannam@0: 
cannam@31: Notes::Notes(float inputSampleRate) :
cannam@0:     Plugin(inputSampleRate),
cannam@0:     m_ibuf(0),
cannam@0:     m_onset(0),
cannam@34:     m_pitch(0),
cannam@0:     m_onsetdet(0),
cannam@30:     m_onsettype(OnsetComplex),
cannam@0:     m_pitchdet(0),
cannam@30:     m_pitchtype(PitchYinFFT),
cannam@0:     m_threshold(0.3),
cannam@34:     m_silence(-70),
cannam@34:     m_minioi(4),
cannam@7:     m_median(6),
cannam@36:     m_minpitch(32),
cannam@7:     m_maxpitch(95),
cannam@7:     m_wrapRange(false),
cannam@7:     m_avoidLeaps(false),
cannam@7:     m_prevPitch(-1)
cannam@0: {
cannam@0: }
cannam@0: 
cannam@0: Notes::~Notes()
cannam@0: {
cannam@30:     if (m_onsetdet) del_aubio_onset(m_onsetdet);
cannam@30:     if (m_pitchdet) del_aubio_pitch(m_pitchdet);
cannam@0:     if (m_ibuf) del_fvec(m_ibuf);
cannam@0:     if (m_onset) del_fvec(m_onset);
cannam@34:     if (m_pitch) del_fvec(m_pitch);
cannam@0: }
cannam@0: 
cannam@0: string
cannam@13: Notes::getIdentifier() const
cannam@0: {
cannam@0:     return "aubionotes";
cannam@0: }
cannam@0: 
cannam@0: string
cannam@13: Notes::getName() const
cannam@13: {
cannam@13:     return "Aubio Note Tracker";
cannam@13: }
cannam@13: 
cannam@13: string
cannam@0: Notes::getDescription() const
cannam@0: {
cannam@13:     return "Estimate note onset positions, pitches and durations";
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: Notes::getMaker() const
cannam@0: {
cannam@0:     return "Paul Brossier (plugin by Chris Cannam)";
cannam@0: }
cannam@0: 
cannam@0: int
cannam@0: Notes::getPluginVersion() const
cannam@0: {
cannam@31:     return 4;
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: Notes::getCopyright() const
cannam@0: {
cannam@0:     return "GPL";
cannam@0: }
cannam@0: 
cannam@0: bool
cannam@0: Notes::initialise(size_t channels, size_t stepSize, size_t blockSize)
cannam@0: {
cannam@32:     if (channels != 1) {
cannam@32:         std::cerr << "Notes::initialise: channels must be 1" << std::endl;
cannam@32:         return false;
cannam@32:     }
cannam@32: 
cannam@0:     m_stepSize = stepSize;
cannam@0:     m_blockSize = blockSize;
cannam@0: 
cannam@32:     m_ibuf = new_fvec(stepSize);
cannam@32:     m_onset = new_fvec(1);
cannam@34:     m_pitch = new_fvec(1);
cannam@0: 
cannam@37:     reset();
cannam@37: 
cannam@37:     return true;
cannam@37: }
cannam@37: 
cannam@37: void
cannam@37: Notes::reset()
cannam@37: {
cannam@37:     if (m_onsetdet) del_aubio_onset(m_onsetdet);
cannam@37:     if (m_pitchdet) del_aubio_pitch(m_pitchdet);
cannam@37: 
cannam@34:     m_onsetdet = new_aubio_onset
cannam@34:         (const_cast<char *>(getAubioNameForOnsetType(m_onsettype)),
cannam@37:          m_blockSize,
cannam@37:          m_stepSize,
cannam@34:          lrintf(m_inputSampleRate));
cannam@34:     
cannam@34:     aubio_onset_set_threshold(m_onsetdet, m_threshold);
cannam@34:     aubio_onset_set_silence(m_onsetdet, m_silence);
cannam@34:     aubio_onset_set_minioi(m_onsetdet, m_minioi);
cannam@0: 
cannam@34:     m_pitchdet = new_aubio_pitch
cannam@34:         (const_cast<char *>(getAubioNameForPitchType(m_pitchtype)),
cannam@37:          m_blockSize,
cannam@37:          m_stepSize,
cannam@34:          lrintf(m_inputSampleRate));
cannam@34: 
cannam@35:     aubio_pitch_set_unit(m_pitchdet, const_cast<char *>("freq"));
cannam@0: 
cannam@0:     m_count = 0;
piem@5:     m_delay = Vamp::RealTime::frame2RealTime((4 + m_median) * m_stepSize,
piem@5:                                        lrintf(m_inputSampleRate));
cannam@0:     m_currentOnset = Vamp::RealTime::zeroTime;
cannam@0:     m_haveCurrent = false;
cannam@7:     m_prevPitch = -1;
cannam@0: }
cannam@0: 
cannam@0: size_t
cannam@0: Notes::getPreferredStepSize() const
cannam@0: {
piem@2:     return 512;
cannam@0: }
cannam@0: 
cannam@0: size_t
cannam@0: Notes::getPreferredBlockSize() const
cannam@0: {
cannam@4:     return 4 * getPreferredStepSize();
cannam@0: }
cannam@0: 
cannam@0: Notes::ParameterList
cannam@0: Notes::getParameterDescriptors() const
cannam@0: {
cannam@0:     ParameterList list;
cannam@0:     
cannam@0:     ParameterDescriptor desc;
cannam@13:     desc.identifier = "onsettype";
cannam@13:     desc.name = "Onset Detection Function Type";
piem@62:     desc.description = "Type of onset detection function to use";
cannam@0:     desc.minValue = 0;
cannam@34:     desc.maxValue = 7;
cannam@34:     desc.defaultValue = (int)OnsetComplex;
cannam@0:     desc.isQuantized = true;
cannam@0:     desc.quantizeStep = 1;
cannam@0:     desc.valueNames.push_back("Energy Based");
cannam@0:     desc.valueNames.push_back("Spectral Difference");
cannam@0:     desc.valueNames.push_back("High-Frequency Content");
cannam@0:     desc.valueNames.push_back("Complex Domain");
cannam@0:     desc.valueNames.push_back("Phase Deviation");
cannam@0:     desc.valueNames.push_back("Kullback-Liebler");
cannam@0:     desc.valueNames.push_back("Modified Kullback-Liebler");
cannam@34:     desc.valueNames.push_back("Spectral Flux");
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "pitchtype";
cannam@13:     desc.name = "Pitch Detection Function Type";
piem@62:     desc.description = "Type of pitch detection function to use";
cannam@0:     desc.minValue = 0;
cannam@0:     desc.maxValue = 4;
cannam@34:     desc.defaultValue = (int)PitchYinFFT;
cannam@0:     desc.isQuantized = true;
cannam@0:     desc.quantizeStep = 1;
cannam@0:     desc.valueNames.push_back("YIN Frequency Estimator");
cannam@0:     desc.valueNames.push_back("Spectral Comb");
cannam@0:     desc.valueNames.push_back("Schmitt");
cannam@0:     desc.valueNames.push_back("Fast Harmonic Comb");
cannam@0:     desc.valueNames.push_back("YIN with FFT");
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "minpitch";
cannam@13:     desc.name = "Minimum Pitch";
piem@78:     desc.description = "Lowest pitch value to look for";
cannam@7:     desc.minValue = 0;
cannam@7:     desc.maxValue = 127;
cannam@36:     desc.defaultValue = 32;
cannam@7:     desc.unit = "MIDI units";
cannam@7:     desc.isQuantized = true;
cannam@7:     desc.quantizeStep = 1;
cannam@7:     list.push_back(desc);
cannam@7: 
cannam@7:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "maxpitch";
cannam@13:     desc.name = "Maximum Pitch";
piem@62:     desc.description = "Highest pitch value to look for";
cannam@7:     desc.minValue = 0;
cannam@7:     desc.maxValue = 127;
cannam@7:     desc.defaultValue = 95;
cannam@7:     desc.unit = "MIDI units";
cannam@7:     desc.isQuantized = true;
cannam@7:     desc.quantizeStep = 1;
cannam@7:     list.push_back(desc);
cannam@7: 
cannam@7:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "wraprange";
cannam@13:     desc.name = "Fold Higher or Lower Notes into Range";
piem@62:     desc.description = "Notes detected outside the range will be transposed to higher or lower octaves";
cannam@7:     desc.minValue = 0;
cannam@7:     desc.maxValue = 1;
cannam@7:     desc.defaultValue = 0;
cannam@7:     desc.isQuantized = true;
cannam@7:     desc.quantizeStep = 1;
cannam@7:     list.push_back(desc);
cannam@7: 
cannam@7:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "avoidleaps";
cannam@13:     desc.name = "Avoid Multi-Octave Jumps";
piem@62:     desc.description = "Minimize octave jumps by transposing to the octave of the previously detected note";
cannam@7:     desc.minValue = 0;
cannam@7:     desc.maxValue = 1;
cannam@7:     desc.defaultValue = 0;
cannam@7:     desc.isQuantized = true;
cannam@7:     desc.quantizeStep = 1;
cannam@7:     list.push_back(desc);
cannam@7: 
cannam@7:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "peakpickthreshold";
cannam@13:     desc.name = "Peak Picker Threshold";
piem@62:     desc.description = "Peak picking threshold, the higher the least detection";
cannam@0:     desc.minValue = 0;
cannam@0:     desc.maxValue = 1;
cannam@0:     desc.defaultValue = 0.3;
cannam@0:     desc.isQuantized = false;
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "silencethreshold";
cannam@13:     desc.name = "Silence Threshold";
piem@62:     desc.description = "Silence threshold, the higher the least detection";
cannam@0:     desc.minValue = -120;
cannam@0:     desc.maxValue = 0;
cannam@34:     desc.defaultValue = -70;
cannam@0:     desc.unit = "dB";
cannam@0:     desc.isQuantized = false;
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@34:     desc = ParameterDescriptor();
cannam@34:     desc.identifier = "minioi";
cannam@34:     desc.name = "Minimum Inter-Onset Interval";
piem@62:     desc.description = "Time interval below which two consecutive onsets should be merged";
cannam@34:     desc.minValue = 0;
cannam@34:     desc.maxValue = 40;
cannam@34:     desc.defaultValue = 4;
cannam@34:     desc.unit = "ms";
cannam@34:     desc.isQuantized = true;
cannam@34:     desc.quantizeStep = 1;
cannam@34:     list.push_back(desc);
cannam@34: 
cannam@0:     return list;
cannam@0: }
cannam@0: 
cannam@0: float
cannam@0: Notes::getParameter(std::string param) const
cannam@0: {
cannam@0:     if (param == "onsettype") {
cannam@0:         return m_onsettype;
cannam@0:     } else if (param == "pitchtype") {
cannam@0:         return m_pitchtype;
cannam@0:     } else if (param == "peakpickthreshold") {
cannam@0:         return m_threshold;
cannam@0:     } else if (param == "silencethreshold") {
cannam@0:         return m_silence;
cannam@7:     } else if (param == "minpitch") {
cannam@7:         return m_minpitch;
cannam@7:     } else if (param == "maxpitch") {
cannam@7:         return m_maxpitch;
cannam@7:     } else if (param == "wraprange") {
cannam@7:         return m_wrapRange ? 1.0 : 0.0;
cannam@7:     } else if (param == "avoidleaps") {
cannam@7:         return m_avoidLeaps ? 1.0 : 0.0;
cannam@34:     } else if (param == "minioi") {
cannam@34:         return m_minioi;
cannam@0:     } else {
cannam@0:         return 0.0;
cannam@0:     }
cannam@0: }
cannam@0: 
cannam@0: void
cannam@0: Notes::setParameter(std::string param, float value)
cannam@0: {
cannam@0:     if (param == "onsettype") {
cannam@0:         switch (lrintf(value)) {
cannam@34:         case 0: m_onsettype = OnsetEnergy; break;
cannam@34:         case 1: m_onsettype = OnsetSpecDiff; break;
cannam@34:         case 2: m_onsettype = OnsetHFC; break;
cannam@34:         case 3: m_onsettype = OnsetComplex; break;
cannam@34:         case 4: m_onsettype = OnsetPhase; break;
cannam@34:         case 5: m_onsettype = OnsetKL; break;
cannam@34:         case 6: m_onsettype = OnsetMKL; break;
cannam@34:         case 7: m_onsettype = OnsetSpecFlux; break;
cannam@0:         }
cannam@0:     } else if (param == "pitchtype") {
cannam@0:         switch (lrintf(value)) {
cannam@34:         case 0: m_pitchtype = PitchYin; break;
cannam@34:         case 1: m_pitchtype = PitchMComb; break;
cannam@34:         case 2: m_pitchtype = PitchSchmitt; break;
cannam@34:         case 3: m_pitchtype = PitchFComb; break;
cannam@34:         case 4: m_pitchtype = PitchYinFFT; break;
cannam@0:         }
cannam@0:     } else if (param == "peakpickthreshold") {
cannam@0:         m_threshold = value;
cannam@0:     } else if (param == "silencethreshold") {
cannam@0:         m_silence = value;
cannam@7:     } else if (param == "minpitch") {
cannam@7:         m_minpitch = lrintf(value);
cannam@7:     } else if (param == "maxpitch") {
cannam@7:         m_maxpitch = lrintf(value);
cannam@7:     } else if (param == "wraprange") {
cannam@7:         m_wrapRange = (value > 0.5);
cannam@7:     } else if (param == "avoidleaps") {
cannam@7:         m_avoidLeaps = (value > 0.5);
cannam@34:     } else if (param == "minioi") {
cannam@34:         m_minioi = value;
cannam@0:     }
cannam@0: }
cannam@0: 
cannam@0: Notes::OutputList
cannam@0: Notes::getOutputDescriptors() const
cannam@0: {
cannam@0:     OutputList list;
cannam@0: 
cannam@0:     OutputDescriptor d;
cannam@13:     d.identifier = "notes";
cannam@13:     d.name = "Notes";
piem@63:     d.description = "List of notes detected, with their frequency and velocity";
cannam@0:     d.unit = "Hz";
cannam@0:     d.hasFixedBinCount = true;
cannam@20: 
cannam@31:     d.binCount = 2;
cannam@31:     d.binNames.push_back("Frequency");
cannam@31:     d.binNames.push_back("Velocity");
cannam@31:     d.hasDuration = true;
cannam@20: 
cannam@0:     d.hasKnownExtents = false;
cannam@0:     d.isQuantized = false;
cannam@0:     d.sampleType = OutputDescriptor::VariableSampleRate;
cannam@0:     d.sampleRate = 0;
cannam@0:     list.push_back(d);
cannam@0: 
cannam@0:     return list;
cannam@0: }
cannam@0: 
cannam@0: Notes::FeatureSet
cannam@12: Notes::process(const float *const *inputBuffers, Vamp::RealTime timestamp)
cannam@0: {
cannam@0:     for (size_t i = 0; i < m_stepSize; ++i) {
piem@52:         fvec_set_sample(m_ibuf, inputBuffers[0][i], i);
cannam@0:     }
cannam@0: 
cannam@34:     aubio_onset_do(m_onsetdet, m_ibuf, m_onset);
cannam@34:     aubio_pitch_do(m_pitchdet, m_ibuf, m_pitch);
cannam@0: 
cannam@34:     bool isonset = m_onset->data[0];
cannam@34:     float frequency = m_pitch->data[0];
cannam@0: 
cannam@0:     m_notebuf.push_back(frequency);
cannam@0:     if (m_notebuf.size() > m_median) m_notebuf.pop_front();
cannam@0: 
cannam@0:     float level = aubio_level_detection(m_ibuf, m_silence);
cannam@0: 
cannam@0:     FeatureSet returnFeatures;
cannam@0: 
cannam@0:     if (isonset) {
cannam@0:         if (level == 1.) {
cannam@0:             isonset = false;
cannam@0:             m_count = 0;
cannam@0:             if (m_haveCurrent) pushNote(returnFeatures, timestamp);
cannam@0:         } else {
cannam@0:             m_count = 1;
cannam@0:         }
cannam@0:     } else {
cannam@0:         if (m_count > 0) ++m_count;
cannam@0:         if (m_count == m_median) {
cannam@0:             if (m_haveCurrent) pushNote(returnFeatures, timestamp);
cannam@0:             m_currentOnset = timestamp;
cannam@0:             m_currentLevel = level;
cannam@0:             m_haveCurrent = true;
cannam@0:         }
cannam@0:     }
cannam@0: 
cannam@0:     m_lastTimeStamp = timestamp;
cannam@0:     return returnFeatures;
cannam@0: }
cannam@0: 
cannam@0: Notes::FeatureSet
cannam@0: Notes::getRemainingFeatures()
cannam@0: {
cannam@0:     FeatureSet returnFeatures;
cannam@0:     if (m_haveCurrent) pushNote(returnFeatures, m_lastTimeStamp);
cannam@0:     return returnFeatures;
cannam@0: }
cannam@0: 
cannam@0: void
cannam@0: Notes::pushNote(FeatureSet &fs, const Vamp::RealTime &offTime)
cannam@0: {
cannam@0:     std::deque<float> toSort = m_notebuf;
cannam@0:     std::sort(toSort.begin(), toSort.end());
cannam@0:     float median = toSort[toSort.size()/2];
cannam@0:     if (median < 45.0) return;
cannam@0: 
cannam@7:     float freq = median;
cannam@15:     int midiPitch = (int)floor(aubio_freqtomidi(freq) + 0.5);
cannam@7:     
cannam@7:     if (m_avoidLeaps) {
cannam@7:         if (m_prevPitch >= 0) {
cannam@7:             while (midiPitch < m_prevPitch - 12) {
cannam@7:                 midiPitch += 12;
cannam@7:                 freq *= 2;
cannam@7:             }
cannam@7:             while (midiPitch > m_prevPitch + 12) {
cannam@7:                 midiPitch -= 12;
cannam@7:                 freq /= 2;
cannam@7:             }
cannam@7:         }
cannam@7:     }
cannam@7: 
cannam@7:     while (midiPitch < m_minpitch) {
cannam@7:         if (!m_wrapRange) return;
cannam@7:         midiPitch += 12;
cannam@7:         freq *= 2;
cannam@7:     }
cannam@7: 
cannam@7:     while (midiPitch > m_maxpitch) {
cannam@7:         if (!m_wrapRange) return;
cannam@7:         midiPitch -= 12;
cannam@7:         freq /= 2;
cannam@7:     }
cannam@7: 
cannam@7:     m_prevPitch = midiPitch;
cannam@7: 
cannam@0:     Feature feature;
cannam@0:     feature.hasTimestamp = true;
piem@5:     if (m_currentOnset < m_delay) m_currentOnset = m_delay;
piem@5:     feature.timestamp = m_currentOnset - m_delay;
cannam@7:     feature.values.push_back(freq);
cannam@20: 
cannam@31:     feature.values.push_back
cannam@31:         (Vamp::RealTime::realTime2Frame
cannam@31:          (offTime, lrintf(m_inputSampleRate)) -
cannam@31:          Vamp::RealTime::realTime2Frame
cannam@31:          (m_currentOnset, lrintf(m_inputSampleRate)));
cannam@31:     feature.hasDuration = false;
cannam@20: 
cannam@0:     feature.values.push_back(m_currentLevel);
cannam@0:     fs[0].push_back(feature);
cannam@0: }
cannam@0: