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@0:     This file copyright 2006 Chris Cannam.
cannam@0:     
cannam@0:     This program is free software; you can redistribute it and/or
cannam@0:     modify it under the terms of the GNU General Public License as
cannam@0:     published by the Free Software Foundation; either version 2 of the
cannam@0:     License, or (at your option) any later version.  See the file
cannam@0:     COPYING included with this distribution for more information.
cannam@0: 
cannam@0: */
cannam@0: 
cannam@0: #include "Notes.h"
cannam@0: 
cannam@0: using std::string;
cannam@0: using std::vector;
cannam@0: using std::cerr;
cannam@0: using std::endl;
cannam@0: 
cannam@0: Notes::Notes(float inputSampleRate) :
cannam@0:     Plugin(inputSampleRate),
cannam@0:     m_ibuf(0),
cannam@0:     m_fftgrain(0),
cannam@0:     m_onset(0),
cannam@0:     m_pv(0),
cannam@0:     m_peakpick(0),
cannam@0:     m_onsetdet(0),
cannam@0:     m_onsettype(aubio_onset_mkl),
cannam@0:     m_pitchdet(0),
cannam@0:     m_pitchtype(aubio_pitch_fcomb),
cannam@0:     m_pitchmode(aubio_pitchm_freq),
cannam@0:     m_threshold(0.3),
cannam@0:     m_silence(-90),
cannam@0:     m_median(6)
cannam@0: {
cannam@0: }
cannam@0: 
cannam@0: Notes::~Notes()
cannam@0: {
cannam@0:     if (m_onsetdet) aubio_onsetdetection_free(m_onsetdet);
cannam@0:     if (m_pitchdet) del_aubio_pitchdetection(m_pitchdet);
cannam@0:     if (m_ibuf) del_fvec(m_ibuf);
cannam@0:     if (m_onset) del_fvec(m_onset);
cannam@0:     if (m_fftgrain) del_cvec(m_fftgrain);
cannam@0:     if (m_pv) del_aubio_pvoc(m_pv);
cannam@0:     if (m_peakpick) del_aubio_peakpicker(m_peakpick);
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: Notes::getName() const
cannam@0: {
cannam@0:     return "aubionotes";
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: Notes::getDescription() const
cannam@0: {
cannam@0:     return "Aubio Note Tracker";
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@0:     return 1;
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@0:     m_channelCount = channels;
cannam@0:     m_stepSize = stepSize;
cannam@0:     m_blockSize = blockSize;
cannam@0: 
cannam@0:     size_t processingBlockSize;
cannam@0:     if (m_onsettype == aubio_onset_energy ||
cannam@0:         m_onsettype == aubio_onset_hfc) {
cannam@0:         processingBlockSize = stepSize * 2;
cannam@0:     } else {
cannam@0:         processingBlockSize = stepSize * 4;
cannam@0:     }
cannam@0: 
cannam@0:     m_ibuf = new_fvec(stepSize, channels);
cannam@0:     m_onset = new_fvec(1, channels);
cannam@0:     m_fftgrain = new_cvec(processingBlockSize, channels);
cannam@0:     m_pv = new_aubio_pvoc(processingBlockSize, stepSize, channels);
cannam@0:     m_peakpick = new_aubio_peakpicker(m_threshold);
cannam@0: 
cannam@0:     m_onsetdet = new_aubio_onsetdetection(m_onsettype, processingBlockSize, channels);
cannam@0: 
cannam@0:     m_pitchdet = new_aubio_pitchdetection(processingBlockSize * 4,
cannam@0:                                           stepSize,
cannam@0:                                           channels,
cannam@0:                                           lrintf(m_inputSampleRate),
cannam@0:                                           m_pitchtype,
cannam@0:                                           m_pitchmode);
cannam@0: 
cannam@0:     m_count = 0;
cannam@0:     m_currentOnset = Vamp::RealTime::zeroTime;
cannam@0:     m_haveCurrent = false;
cannam@0: 
cannam@0:     return true;
cannam@0: }
cannam@0: 
cannam@0: void
cannam@0: Notes::reset()
cannam@0: {
cannam@0: }
cannam@0: 
cannam@0: size_t
cannam@0: Notes::getPreferredStepSize() const
cannam@0: {
cannam@0:     if (m_onsettype == aubio_onset_energy ||
cannam@0:         m_onsettype == aubio_onset_hfc) {
cannam@0:         return 512;
cannam@0:     } else {
cannam@0:         return 128;
cannam@0:     }
cannam@0: }
cannam@0: 
cannam@0: size_t
cannam@0: Notes::getPreferredBlockSize() const
cannam@0: {
cannam@0:     return 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@0:     desc.name = "onsettype";
cannam@0:     desc.description = "Onset Detection Function Type";
cannam@0:     desc.minValue = 0;
cannam@0:     desc.maxValue = 6;
cannam@0:     desc.defaultValue = (int)aubio_onset_mkl;
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@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc = ParameterDescriptor();
cannam@0:     desc.name = "pitchtype";
cannam@0:     desc.description = "Pitch Detection Function Type";
cannam@0:     desc.minValue = 0;
cannam@0:     desc.maxValue = 4;
cannam@0:     desc.defaultValue = (int)aubio_pitch_fcomb;
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@0:     desc.name = "peakpickthreshold";
cannam@0:     desc.description = "Peak Picker Threshold";
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@0:     desc.name = "silencethreshold";
cannam@0:     desc.description = "Silence Threshold";
cannam@0:     desc.minValue = -120;
cannam@0:     desc.maxValue = 0;
cannam@0:     desc.defaultValue = -90;
cannam@0:     desc.unit = "dB";
cannam@0:     desc.isQuantized = false;
cannam@0:     list.push_back(desc);
cannam@0: 
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@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@0:         case 0: m_onsettype = aubio_onset_energy; break;
cannam@0:         case 1: m_onsettype = aubio_onset_specdiff; break;
cannam@0:         case 2: m_onsettype = aubio_onset_hfc; break;
cannam@0:         case 3: m_onsettype = aubio_onset_complex; break;
cannam@0:         case 4: m_onsettype = aubio_onset_phase; break;
cannam@0:         case 5: m_onsettype = aubio_onset_kl; break;
cannam@0:         case 6: m_onsettype = aubio_onset_mkl; break;
cannam@0:         }
cannam@0:     } else if (param == "pitchtype") {
cannam@0:         switch (lrintf(value)) {
cannam@0:         case 0: m_pitchtype = aubio_pitch_yin; break;
cannam@0:         case 1: m_pitchtype = aubio_pitch_mcomb; break;
cannam@0:         case 2: m_pitchtype = aubio_pitch_schmitt; break;
cannam@0:         case 3: m_pitchtype = aubio_pitch_fcomb; break;
cannam@0:         case 4: m_pitchtype = aubio_pitch_yinfft; 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@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@0:     d.name = "notes";
cannam@0:     d.unit = "Hz";
cannam@0:     d.description = "Notes";
cannam@0:     d.hasFixedBinCount = true;
cannam@0:     d.binCount = 2;
cannam@0:     d.binNames.push_back("Frequency");
cannam@0:     d.binNames.push_back("Duration");
cannam@0:     d.binNames.push_back("Velocity");
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@0: Notes::process(float **inputBuffers, Vamp::RealTime timestamp)
cannam@0: {
cannam@0:     for (size_t i = 0; i < m_stepSize; ++i) {
cannam@0:         for (size_t j = 0; j < m_channelCount; ++j) {
cannam@0:             fvec_write_sample(m_ibuf, inputBuffers[j][i], j, i);
cannam@0:         }
cannam@0:     }
cannam@0: 
cannam@0:     aubio_pvoc_do(m_pv, m_ibuf, m_fftgrain);
cannam@0:     aubio_onsetdetection(m_onsetdet, m_fftgrain, m_onset);
cannam@0: 
cannam@0:     bool isonset = aubio_peakpick_pimrt(m_onset, m_peakpick);
cannam@0: 
cannam@0:     float frequency = aubio_pitchdetection(m_pitchdet, m_ibuf);
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@0:     Feature feature;
cannam@0:     feature.hasTimestamp = true;
cannam@0:     feature.timestamp = m_currentOnset;
cannam@0:     feature.values.push_back(median);
cannam@0: //    feature.values.push_back(FLOOR(aubio_freqtomidi(median) + 0.5));
cannam@0:     feature.values.push_back
cannam@0:         (Vamp::RealTime::realTime2Frame(offTime, lrintf(m_inputSampleRate)) -
cannam@0:          Vamp::RealTime::realTime2Frame(m_currentOnset, lrintf(m_inputSampleRate)));
cannam@0:     feature.values.push_back(m_currentLevel);
cannam@0:     fs[0].push_back(feature);
cannam@0: }
cannam@0: