cannam@7: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
cannam@7: 
cannam@7: /*
cannam@7:     Vamp feature extraction plugins using Paul Brossier's Aubio library.
cannam@7: 
cannam@7:     Centre for Digital Music, Queen Mary, University of London.
cannam@7:     This file copyright 2006 Chris Cannam.
cannam@7:     
cannam@7:     This program is free software; you can redistribute it and/or
cannam@7:     modify it under the terms of the GNU General Public License as
cannam@7:     published by the Free Software Foundation; either version 2 of the
cannam@7:     License, or (at your option) any later version.  See the file
cannam@7:     COPYING included with this distribution for more information.
cannam@7: 
cannam@7: */
cannam@7: 
cannam@7: #include <math.h>
cannam@7: #include "Tempo.h"
cannam@7: 
cannam@7: using std::string;
cannam@7: using std::vector;
cannam@7: using std::cerr;
cannam@7: using std::endl;
cannam@7: 
cannam@14: //#define HAVE_AUBIO_LOCKED_TEMPO_HACK
cannam@12: 
cannam@7: Tempo::Tempo(float inputSampleRate) :
cannam@7:     Plugin(inputSampleRate),
cannam@7:     m_ibuf(0),
cannam@7:     m_fftgrain(0),
cannam@7:     m_onset(0),
cannam@7:     m_pv(0),
cannam@7:     m_peakpick(0),
cannam@7:     m_onsetdet(0),
cannam@7:     m_onsettype(aubio_onset_specdiff),
cannam@7:     m_beattracking(0),
cannam@7:     m_dfframe(0),
cannam@7:     m_btout(0),
cannam@7:     m_btcounter(0),
cannam@7:     m_threshold(0.3),
cannam@7:     m_silence(-90),
cannam@7:     m_channelCount(1)
cannam@7: {
cannam@7: }
cannam@7: 
cannam@7: Tempo::~Tempo()
cannam@7: {
cannam@7:     if (m_onsetdet) aubio_onsetdetection_free(m_onsetdet);
cannam@7:     if (m_ibuf) del_fvec(m_ibuf);
cannam@7:     if (m_onset) del_fvec(m_onset);
cannam@7:     if (m_fftgrain) del_cvec(m_fftgrain);
cannam@7:     if (m_pv) del_aubio_pvoc(m_pv);
cannam@7:     if (m_peakpick) del_aubio_peakpicker(m_peakpick);
cannam@7:     if (m_beattracking) del_aubio_beattracking(m_beattracking);
cannam@7:     if (m_dfframe) del_fvec(m_dfframe);
cannam@7:     if (m_btout) del_fvec(m_btout);
cannam@7: }
cannam@7: 
cannam@7: string
cannam@13: Tempo::getIdentifier() const
cannam@7: {
cannam@7:     return "aubiotempo";
cannam@7: }
cannam@7: 
cannam@7: string
cannam@13: Tempo::getName() const
cannam@7: {
cannam@7:     return "Aubio Tempo Detector";
cannam@7: }
cannam@7: 
cannam@7: string
cannam@13: Tempo::getDescription() const
cannam@13: {
cannam@13:     return "Estimate the musical tempo by tracking note onset timings";
cannam@13: }
cannam@13: 
cannam@13: string
cannam@7: Tempo::getMaker() const
cannam@7: {
cannam@13:     return "Paul Brossier (method by Matthew Davies, plugin by Chris Cannam)";
cannam@7: }
cannam@7: 
cannam@7: int
cannam@7: Tempo::getPluginVersion() const
cannam@7: {
cannam@7:     return 1;
cannam@7: }
cannam@7: 
cannam@7: string
cannam@7: Tempo::getCopyright() const
cannam@7: {
cannam@7:     return "GPL";
cannam@7: }
cannam@7: 
cannam@7: bool
cannam@7: Tempo::initialise(size_t channels, size_t stepSize, size_t blockSize)
cannam@7: {
cannam@7:     m_channelCount = channels;
cannam@7:     m_stepSize = stepSize;
cannam@7:     m_blockSize = blockSize;
cannam@7: 
cannam@7:     m_ibuf = new_fvec(stepSize, channels);
cannam@7:     m_onset = new_fvec(1, channels);
cannam@7:     m_fftgrain = new_cvec(blockSize, channels);
cannam@7:     m_pv = new_aubio_pvoc(blockSize, stepSize, channels);
cannam@7:     m_peakpick = new_aubio_peakpicker(m_threshold);
cannam@7: 
cannam@7:     m_onsetdet = new_aubio_onsetdetection(m_onsettype, blockSize, channels);
cannam@7:     
cannam@7:     m_delay = Vamp::RealTime::frame2RealTime(3 * stepSize,
cannam@7:                                              lrintf(m_inputSampleRate));
cannam@7: 
cannam@7:     m_lastBeat = Vamp::RealTime::zeroTime - m_delay - m_delay;
cannam@7: 
cannam@7:     m_winlen = 512*512/stepSize;
cannam@7:     m_dfframe = new_fvec(m_winlen,channels);
cannam@7:     m_btstep = m_winlen/4;
cannam@7:     m_btout = new_fvec(m_btstep,channels);
cannam@7:     m_beattracking = new_aubio_beattracking(m_winlen,channels);
cannam@7: 
cannam@7:     return true;
cannam@7: }
cannam@7: 
cannam@7: void
cannam@7: Tempo::reset()
cannam@7: {
cannam@7: }
cannam@7: 
cannam@7: size_t
cannam@7: Tempo::getPreferredStepSize() const
cannam@7: {
cannam@7:     return 512;
cannam@7: }
cannam@7: 
cannam@7: size_t
cannam@7: Tempo::getPreferredBlockSize() const
cannam@7: {
cannam@7:     return 2 * getPreferredStepSize();
cannam@7: }
cannam@7: 
cannam@7: Tempo::ParameterList
cannam@7: Tempo::getParameterDescriptors() const
cannam@7: {
cannam@7:     ParameterList list;
cannam@7:     
cannam@7:     ParameterDescriptor desc;
cannam@13:     desc.identifier = "onsettype";
cannam@13:     desc.name = "Onset Detection Function Type";
cannam@7:     desc.minValue = 0;
cannam@7:     desc.maxValue = 6;
cannam@7:     desc.defaultValue = (int)aubio_onset_complex;
cannam@7:     desc.isQuantized = true;
cannam@7:     desc.quantizeStep = 1;
cannam@7:     desc.valueNames.push_back("Energy Based");
cannam@7:     desc.valueNames.push_back("Spectral Difference");
cannam@7:     desc.valueNames.push_back("High-Frequency Content");
cannam@7:     desc.valueNames.push_back("Complex Domain");
cannam@7:     desc.valueNames.push_back("Phase Deviation");
cannam@7:     desc.valueNames.push_back("Kullback-Liebler");
cannam@7:     desc.valueNames.push_back("Modified Kullback-Liebler");
cannam@7:     list.push_back(desc);
cannam@7: 
cannam@7:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "peakpickthreshold";
cannam@13:     desc.name = "Peak Picker Threshold";
cannam@7:     desc.minValue = 0;
cannam@7:     desc.maxValue = 1;
cannam@7:     desc.defaultValue = 0.3;
cannam@7:     desc.isQuantized = false;
cannam@7:     list.push_back(desc);
cannam@7: 
cannam@7:     desc = ParameterDescriptor();
cannam@13:     desc.identifier = "silencethreshold";
cannam@13:     desc.name = "Silence Threshold";
cannam@7:     desc.minValue = -120;
cannam@7:     desc.maxValue = 0;
cannam@7:     desc.defaultValue = -90;
cannam@7:     desc.unit = "dB";
cannam@7:     desc.isQuantized = false;
cannam@7:     list.push_back(desc);
cannam@7: 
cannam@7:     return list;
cannam@7: }
cannam@7: 
cannam@7: float
cannam@7: Tempo::getParameter(std::string param) const
cannam@7: {
cannam@7:     if (param == "onsettype") {
cannam@7:         return m_onsettype;
cannam@7:     } else if (param == "peakpickthreshold") {
cannam@7:         return m_threshold;
cannam@7:     } else if (param == "silencethreshold") {
cannam@7:         return m_silence;
cannam@7:     } else {
cannam@7:         return 0.0;
cannam@7:     }
cannam@7: }
cannam@7: 
cannam@7: void
cannam@7: Tempo::setParameter(std::string param, float value)
cannam@7: {
cannam@7:     if (param == "onsettype") {
cannam@7:         switch (lrintf(value)) {
cannam@7:         case 0: m_onsettype = aubio_onset_energy; break;
cannam@7:         case 1: m_onsettype = aubio_onset_specdiff; break;
cannam@7:         case 2: m_onsettype = aubio_onset_hfc; break;
cannam@7:         case 3: m_onsettype = aubio_onset_complex; break;
cannam@7:         case 4: m_onsettype = aubio_onset_phase; break;
cannam@7:         case 5: m_onsettype = aubio_onset_kl; break;
cannam@7:         case 6: m_onsettype = aubio_onset_mkl; break;
cannam@7:         }
cannam@7:     } else if (param == "peakpickthreshold") {
cannam@7:         m_threshold = value;
cannam@7:     } else if (param == "silencethreshold") {
cannam@7:         m_silence = value;
cannam@7:     }
cannam@7: }
cannam@7: 
cannam@7: Tempo::OutputList
cannam@7: Tempo::getOutputDescriptors() const
cannam@7: {
cannam@7:     OutputList list;
cannam@7: 
cannam@7:     OutputDescriptor d;
cannam@13:     d.identifier = "beats";
cannam@13:     d.name = "Beats";
cannam@7:     d.unit = "";
cannam@7:     d.hasFixedBinCount = true;
cannam@7:     d.binCount = 0;
cannam@7:     d.sampleType = OutputDescriptor::VariableSampleRate;
cannam@7:     d.sampleRate = 0;
cannam@7:     list.push_back(d);
cannam@7: 
cannam@12: #ifdef HAVE_AUBIO_LOCKED_TEMPO_HACK
cannam@13:     d.identifier = "tempo";
cannam@13:     d.name = "Tempo";
cannam@12:     d.unit = "bpm";
cannam@12:     d.hasFixedBinCount = true;
cannam@12:     d.binCount = 1;
cannam@12:     d.hasKnownExtents = false;
cannam@12:     d.isQuantized = false;
cannam@12:     d.sampleType = OutputDescriptor::OneSamplePerStep;
cannam@12:     list.push_back(d);
cannam@12: #endif
cannam@12: 
cannam@7:     return list;
cannam@7: }
cannam@7: 
cannam@7: Tempo::FeatureSet
cannam@12: Tempo::process(const float *const *inputBuffers, Vamp::RealTime timestamp)
cannam@7: {
cannam@7:     for (size_t i = 0; i < m_stepSize; ++i) {
cannam@7:         for (size_t j = 0; j < m_channelCount; ++j) {
cannam@7:             fvec_write_sample(m_ibuf, inputBuffers[j][i], j, i);
cannam@7:         }
cannam@7:     }
cannam@7: 
cannam@7:     aubio_pvoc_do(m_pv, m_ibuf, m_fftgrain);
cannam@7:     aubio_onsetdetection(m_onsetdet, m_fftgrain, m_onset);
cannam@7: 
cannam@12: #ifdef HAVE_AUBIO_LOCKED_TEMPO_HACK
cannam@12:     float locked_tempo = 0;
cannam@12: #endif
cannam@12: 
cannam@7:     if ( m_btcounter == m_btstep - 1 ) {
cannam@12: #ifdef HAVE_AUBIO_LOCKED_TEMPO_HACK
cannam@12:         aubio_beattracking_do(m_beattracking,m_dfframe,m_btout,&locked_tempo);
cannam@12: #else
cannam@7:         aubio_beattracking_do(m_beattracking,m_dfframe,m_btout);
cannam@12: #endif
cannam@7:         /* rotate dfframe */
cannam@7:         for (size_t i = 0 ; i < m_winlen - m_btstep; i++ ) 
cannam@7:                 m_dfframe->data[0][i] = m_dfframe->data[0][i+m_btstep];
cannam@7:         for (size_t i = m_winlen - m_btstep ; i < m_winlen; i++ ) 
cannam@7:                 m_dfframe->data[0][i] = 0.;
cannam@7:                 
cannam@7:         m_btcounter = -1;
cannam@7:     }
cannam@7:     m_btcounter++;
cannam@7:     bool isonset = aubio_peakpick_pimrt_wt( m_onset, m_peakpick, 
cannam@7:         &(m_dfframe->data[0][m_winlen - m_btstep + m_btcounter]));
cannam@7:     bool istactus = 0;
cannam@7: 
cannam@7:     /* check if any of the predicted beat correspond to the current time */
cannam@7:     for (size_t i = 1; i < m_btout->data[0][0]; i++ ) { 
cannam@7:             if (m_btcounter == m_btout->data[0][i]) {
cannam@7:                     if (aubio_silence_detection(m_ibuf, m_silence)) {
cannam@7:                             isonset  = false;
cannam@7:                             istactus = false;
cannam@7:                     } else {
cannam@7:                             istactus = true;
cannam@7:                     }
cannam@7:             }
cannam@7:     }
cannam@7: 
cannam@7:     FeatureSet returnFeatures;
cannam@7: 
cannam@7:     if (istactus == true) {
cannam@7:         if (timestamp - m_lastBeat >= m_delay) {
cannam@7:             Feature onsettime;
cannam@7:             onsettime.hasTimestamp = true;
cannam@7:             if (timestamp < m_delay) timestamp = m_delay;
cannam@7:             onsettime.timestamp = timestamp - m_delay;
cannam@7:             returnFeatures[0].push_back(onsettime);
cannam@7:             m_lastBeat = timestamp;
cannam@7:         }
cannam@7:     }
cannam@7: 
cannam@12: #ifdef HAVE_AUBIO_LOCKED_TEMPO_HACK
cannam@12:     if (locked_tempo >= 30 && locked_tempo <= 206) {
cannam@12:         if (locked_tempo > 145) locked_tempo /= 2;
cannam@12:         std::cerr << "Locked tempo: " << locked_tempo << std::endl;
cannam@12:         Feature tempo;
cannam@12:         tempo.hasTimestamp = false;
cannam@12:         tempo.values.push_back(locked_tempo);
cannam@12:         returnFeatures[1].push_back(tempo);
cannam@12:     }
cannam@12: #endif
cannam@12: 
cannam@7:     return returnFeatures;
cannam@7: }
cannam@7: 
cannam@7: Tempo::FeatureSet
cannam@7: Tempo::getRemainingFeatures()
cannam@7: {
cannam@7:     return FeatureSet();
cannam@7: }
cannam@7: