Chris@23: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
matthiasm@0: 
matthiasm@0: #include "NNLSChroma.h"
Chris@27: 
Chris@27: #include "chromamethods.h"
Chris@27: 
Chris@27: #include <cstdlib>
Chris@27: #include <fstream>
matthiasm@0: #include <cmath>
matthiasm@9: 
Chris@27: #include <algorithm>
matthiasm@0: 
matthiasm@0: const bool debug_on = false;
matthiasm@0: 
Chris@27: const vector<float> hw(hammingwind, hammingwind+19);
matthiasm@0: 
matthiasm@0: NNLSChroma::NNLSChroma(float inputSampleRate) :
Chris@23:     Plugin(inputSampleRate),
Chris@23:     m_fl(0),
Chris@23:     m_blockSize(0),
Chris@23:     m_stepSize(0),
Chris@23:     m_lengthOfNoteIndex(0),
Chris@23:     m_meanTuning0(0),
Chris@23:     m_meanTuning1(0),
Chris@23:     m_meanTuning2(0),
Chris@23:     m_localTuning0(0),
Chris@23:     m_localTuning1(0),
Chris@23:     m_localTuning2(0),
Chris@23:     m_paling(1.0),
Chris@23:     m_preset(0.0),
Chris@23:     m_localTuning(0),
Chris@23:     m_kernelValue(0),
Chris@23:     m_kernelFftIndex(0),
Chris@23:     m_kernelNoteIndex(0),
Chris@23:     m_dict(0),
Chris@23:     m_tuneLocal(false),
Chris@23:     m_dictID(0),
Chris@23:     m_chorddict(0),
Chris@23:     m_chordnames(0),
Chris@23:     m_doNormalizeChroma(0),
Chris@23:     m_rollon(0.01)
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> NNLSChroma" << endl;
matthiasm@7: 
Chris@23:     // make the *note* dictionary matrix
Chris@23:     m_dict = new float[nNote * 84];
Chris@23:     for (unsigned i = 0; i < nNote * 84; ++i) m_dict[i] = 0.0;
Chris@23:     dictionaryMatrix(m_dict);
matthiasm@7: 	
Chris@23:     // get the *chord* dictionary from file (if the file exists)
Chris@23:     m_chordnames = chordDictionary(&m_chorddict);
matthiasm@0: }
matthiasm@0: 
matthiasm@0: 
matthiasm@0: NNLSChroma::~NNLSChroma()
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> ~NNLSChroma" << endl;
Chris@23:     delete [] m_dict;
Chris@23:     // delete [] m_chorddict;
Chris@23:     // delete m_chordnames;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: string
matthiasm@0: NNLSChroma::getIdentifier() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getIdentifier" << endl;
matthiasm@0:     return "nnls_chroma";
matthiasm@0: }
matthiasm@0: 
matthiasm@0: string
matthiasm@0: NNLSChroma::getName() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getName" << endl;
matthiasm@0:     return "NNLS Chroma";
matthiasm@0: }
matthiasm@0: 
matthiasm@0: string
matthiasm@0: NNLSChroma::getDescription() const
matthiasm@0: {
matthiasm@0:     // Return something helpful here!
Chris@23:     if (debug_on) cerr << "--> getDescription" << endl;
matthiasm@13:     return "This plugin provides a number of features derived from a log-frequency amplitude spectrum of the DFT: some variants of the log-frequency spectrum, including a semitone spectrum derived from approximate transcription using the NNLS algorithm; based on this semitone spectrum, chroma features and a simple chord estimate.";
matthiasm@0: }
matthiasm@0: 
matthiasm@0: string
matthiasm@0: NNLSChroma::getMaker() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getMaker" << endl;
matthiasm@0:     // Your name here
matthiasm@0:     return "Matthias Mauch";
matthiasm@0: }
matthiasm@0: 
matthiasm@0: int
matthiasm@0: NNLSChroma::getPluginVersion() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getPluginVersion" << endl;
matthiasm@0:     // Increment this each time you release a version that behaves
matthiasm@0:     // differently from the previous one
matthiasm@0:     return 1;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: string
matthiasm@0: NNLSChroma::getCopyright() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getCopyright" << endl;
matthiasm@0:     // This function is not ideally named.  It does not necessarily
matthiasm@0:     // need to say who made the plugin -- getMaker does that -- but it
matthiasm@0:     // should indicate the terms under which it is distributed.  For
matthiasm@0:     // example, "Copyright (year). All Rights Reserved", or "GPL"
matthiasm@0:     return "Copyright (2010). All rights reserved.";
matthiasm@0: }
matthiasm@0: 
matthiasm@0: NNLSChroma::InputDomain
matthiasm@0: NNLSChroma::getInputDomain() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getInputDomain" << endl;
matthiasm@0:     return FrequencyDomain;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: size_t
matthiasm@0: NNLSChroma::getPreferredBlockSize() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getPreferredBlockSize" << endl;
matthiasm@0:     return 16384; // 0 means "I can handle any block size"
matthiasm@0: }
matthiasm@0: 
matthiasm@0: size_t 
matthiasm@0: NNLSChroma::getPreferredStepSize() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getPreferredStepSize" << endl;
matthiasm@0:     return 2048; // 0 means "anything sensible"; in practice this
Chris@23:     // means the same as the block size for TimeDomain
Chris@23:     // plugins, or half of it for FrequencyDomain plugins
matthiasm@0: }
matthiasm@0: 
matthiasm@0: size_t
matthiasm@0: NNLSChroma::getMinChannelCount() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getMinChannelCount" << endl;
matthiasm@0:     return 1;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: size_t
matthiasm@0: NNLSChroma::getMaxChannelCount() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getMaxChannelCount" << endl;
matthiasm@0:     return 1;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: NNLSChroma::ParameterList
matthiasm@0: NNLSChroma::getParameterDescriptors() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getParameterDescriptors" << endl;
matthiasm@0:     ParameterList list;
matthiasm@0: 
matthiasm@3:     ParameterDescriptor d3;
matthiasm@3:     d3.identifier = "preset";
matthiasm@3:     d3.name = "preset";
matthiasm@3:     d3.description = "Spectral paling: no paling - 0; whitening - 1.";
matthiasm@3:     d3.unit = "";
Chris@23:     d3.isQuantized = true;
Chris@23:     d3.quantizeStep = 1;
matthiasm@3:     d3.minValue = 0.0;
matthiasm@4:     d3.maxValue = 3.0;
matthiasm@3:     d3.defaultValue = 0.0;
matthiasm@3:     d3.valueNames.push_back("polyphonic pop");
Chris@23:     d3.valueNames.push_back("polyphonic pop (fast)");
matthiasm@3:     d3.valueNames.push_back("solo keyboard");
Chris@23:     d3.valueNames.push_back("manual");
matthiasm@3:     list.push_back(d3);
matthiasm@4: 
matthiasm@17:     ParameterDescriptor d5;
Chris@23:     d5.identifier = "rollon";
Chris@23:     d5.name = "spectral roll-on";
Chris@23:     d5.description = "The bins below the spectral roll-on quantile will be set to 0.";
Chris@23:     d5.unit = "";
Chris@23:     d5.minValue = 0;
Chris@23:     d5.maxValue = 1;
Chris@23:     d5.defaultValue = 0;
Chris@23:     d5.isQuantized = false;
Chris@23:     list.push_back(d5);
matthiasm@17: 
matthiasm@4:     // ParameterDescriptor d0;
matthiasm@4:     //  d0.identifier = "notedict";
matthiasm@4:     //  d0.name = "note dictionary";
matthiasm@4:     //  d0.description = "Notes in different note dictionaries differ by their spectral shapes.";
matthiasm@4:     //  d0.unit = "";
matthiasm@4:     //  d0.minValue = 0;
matthiasm@4:     //  d0.maxValue = 1;
matthiasm@4:     //  d0.defaultValue = 0;
matthiasm@4:     //  d0.isQuantized = true;
matthiasm@4:     //  d0.valueNames.push_back("s = 0.6");
matthiasm@4:     //  d0.valueNames.push_back("no NNLS");
matthiasm@4:     //  d0.quantizeStep = 1.0;
matthiasm@4:     //  list.push_back(d0);
matthiasm@4: 
matthiasm@4:     ParameterDescriptor d1;
matthiasm@4:     d1.identifier = "tuningmode";
matthiasm@4:     d1.name = "tuning mode";
matthiasm@4:     d1.description = "Tuning can be performed locally or on the whole extraction segment. Local tuning is only advisable when the tuning is likely to change over the audio, for example in podcasts, or in a cappella singing.";
matthiasm@4:     d1.unit = "";
matthiasm@4:     d1.minValue = 0;
matthiasm@4:     d1.maxValue = 1;
matthiasm@4:     d1.defaultValue = 0;
matthiasm@4:     d1.isQuantized = true;
matthiasm@4:     d1.valueNames.push_back("global tuning");
matthiasm@4:     d1.valueNames.push_back("local tuning");
matthiasm@4:     d1.quantizeStep = 1.0;
matthiasm@4:     list.push_back(d1);
matthiasm@4: 
Chris@23:     //     ParameterDescriptor d2;
Chris@23:     //     d2.identifier = "paling";
Chris@23:     //     d2.name = "spectral paling";
Chris@23:     //     d2.description = "Spectral paling: no paling - 0; whitening - 1.";
Chris@23:     //     d2.unit = "";
Chris@23:     // d2.isQuantized = true;
Chris@23:     // // d2.quantizeStep = 0.1;
Chris@23:     //     d2.minValue = 0.0;
Chris@23:     //     d2.maxValue = 1.0;
Chris@23:     //     d2.defaultValue = 1.0;
Chris@23:     //     d2.isQuantized = false;
Chris@23:     //     list.push_back(d2);
Chris@23:     ParameterDescriptor d4;
matthiasm@12:     d4.identifier = "chromanormalize";
matthiasm@12:     d4.name = "chroma normalization";
matthiasm@12:     d4.description = "How shall the chroma vector be normalized?";
matthiasm@12:     d4.unit = "";
matthiasm@12:     d4.minValue = 0;
matthiasm@13:     d4.maxValue = 3;
matthiasm@12:     d4.defaultValue = 0;
matthiasm@12:     d4.isQuantized = true;
matthiasm@13:     d4.valueNames.push_back("none");
matthiasm@13:     d4.valueNames.push_back("maximum norm");
Chris@23:     d4.valueNames.push_back("L1 norm");
Chris@23:     d4.valueNames.push_back("L2 norm");
matthiasm@12:     d4.quantizeStep = 1.0;
matthiasm@12:     list.push_back(d4);
matthiasm@4: 
matthiasm@0:     return list;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: float
matthiasm@0: NNLSChroma::getParameter(string identifier) const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getParameter" << endl;
matthiasm@0:     if (identifier == "notedict") {
matthiasm@0:         return m_dictID; 
matthiasm@0:     }
matthiasm@0:     
matthiasm@0:     if (identifier == "paling") {
matthiasm@0:         return m_paling; 
matthiasm@0:     }
matthiasm@17: 
Chris@23:     if (identifier == "rollon") {
matthiasm@17:         return m_rollon; 
matthiasm@17:     }
matthiasm@0:     
matthiasm@0:     if (identifier == "tuningmode") {
matthiasm@0:         if (m_tuneLocal) {
matthiasm@0:             return 1.0;
matthiasm@0:         } else {
matthiasm@0:             return 0.0;
matthiasm@0:         }
matthiasm@0:     }
Chris@23:     if (identifier == "preset") {
Chris@23:         return m_preset;
matthiasm@3:     }
Chris@23:     if (identifier == "chromanormalize") {
Chris@23:         return m_doNormalizeChroma;
matthiasm@12:     }
matthiasm@0:     return 0;
matthiasm@0:     
matthiasm@0: }
matthiasm@0: 
matthiasm@0: void
matthiasm@0: NNLSChroma::setParameter(string identifier, float value) 
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> setParameter" << endl;
matthiasm@0:     if (identifier == "notedict") {
matthiasm@0:         m_dictID = (int) value;
matthiasm@0:     }
matthiasm@0:     
matthiasm@0:     if (identifier == "paling") {
matthiasm@0:         m_paling = value;
matthiasm@0:     }
matthiasm@0:     
matthiasm@0:     if (identifier == "tuningmode") {
matthiasm@0:         m_tuneLocal = (value > 0) ? true : false;
matthiasm@0:         // cerr << "m_tuneLocal :" << m_tuneLocal << endl;
matthiasm@0:     }
matthiasm@3:     if (identifier == "preset") {
matthiasm@3:         m_preset = value;
Chris@23:         if (m_preset == 0.0) {
Chris@23:             m_tuneLocal = false;
Chris@23:             m_paling = 1.0;
Chris@23:             m_dictID = 0.0;
Chris@23:         }
Chris@23:         if (m_preset == 1.0) {
Chris@23:             m_tuneLocal = false;
Chris@23:             m_paling = 1.0;
Chris@23:             m_dictID = 1.0;
Chris@23:         }
Chris@23:         if (m_preset == 2.0) {
Chris@23:             m_tuneLocal = false;
Chris@23:             m_paling = 0.7;
Chris@23:             m_dictID = 0.0;
Chris@23:         }
matthiasm@3:     }
Chris@23:     if (identifier == "chromanormalize") {
Chris@23:         m_doNormalizeChroma = value;
Chris@23:     }
matthiasm@17: 	 
Chris@23:     if (identifier == "rollon") {
Chris@23:         m_rollon = value;
Chris@23:     }
matthiasm@0: }
matthiasm@0: 
matthiasm@0: NNLSChroma::ProgramList
matthiasm@0: NNLSChroma::getPrograms() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getPrograms" << endl;
matthiasm@0:     ProgramList list;
matthiasm@0: 
matthiasm@0:     // If you have no programs, return an empty list (or simply don't
matthiasm@0:     // implement this function or getCurrentProgram/selectProgram)
matthiasm@0: 
matthiasm@0:     return list;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: string
matthiasm@0: NNLSChroma::getCurrentProgram() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getCurrentProgram" << endl;
matthiasm@0:     return ""; // no programs
matthiasm@0: }
matthiasm@0: 
matthiasm@0: void
matthiasm@0: NNLSChroma::selectProgram(string name)
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> selectProgram" << endl;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: 
matthiasm@0: NNLSChroma::OutputList
matthiasm@0: NNLSChroma::getOutputDescriptors() const
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getOutputDescriptors" << endl;
matthiasm@0:     OutputList list;
matthiasm@0:     
matthiasm@0:     // Make chroma names for the binNames property
matthiasm@0:     vector<string> chromanames;
matthiasm@0:     vector<string> bothchromanames;
matthiasm@0:     for (int iNote = 0; iNote < 24; iNote++) {
matthiasm@0:         bothchromanames.push_back(notenames[iNote]);
matthiasm@0:         if (iNote < 12) {
matthiasm@0:             chromanames.push_back(notenames[iNote]);
matthiasm@0:         }
matthiasm@0:     }
matthiasm@0:     
Chris@23:     // int nNote = 84;
matthiasm@0: 
matthiasm@0:     // See OutputDescriptor documentation for the possibilities here.
matthiasm@0:     // Every plugin must have at least one output.
matthiasm@0: 
matthiasm@0:     OutputDescriptor d0;
matthiasm@0:     d0.identifier = "tuning";
matthiasm@0:     d0.name = "Tuning";
matthiasm@0:     d0.description = "The concert pitch.";
matthiasm@0:     d0.unit = "Hz";
matthiasm@0:     d0.hasFixedBinCount = true;
matthiasm@0:     d0.binCount = 0;
matthiasm@0:     d0.hasKnownExtents = true;
Chris@23:     d0.minValue = 427.47;
Chris@23:     d0.maxValue = 452.89;
matthiasm@0:     d0.isQuantized = false;
matthiasm@0:     d0.sampleType = OutputDescriptor::VariableSampleRate;
matthiasm@0:     d0.hasDuration = false;
matthiasm@0:     list.push_back(d0);
matthiasm@0: 	
Chris@23:     OutputDescriptor d1;
matthiasm@0:     d1.identifier = "logfreqspec";
matthiasm@0:     d1.name = "Log-Frequency Spectrum";
matthiasm@0:     d1.description = "A Log-Frequency Spectrum (constant Q) that is obtained by cosine filter mapping.";
matthiasm@0:     d1.unit = "";
matthiasm@0:     d1.hasFixedBinCount = true;
matthiasm@0:     d1.binCount = nNote;
matthiasm@0:     d1.hasKnownExtents = false;
matthiasm@0:     d1.isQuantized = false;
matthiasm@0:     d1.sampleType = OutputDescriptor::FixedSampleRate;
matthiasm@0:     d1.hasDuration = false;
matthiasm@0:     d1.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@0:     list.push_back(d1);
matthiasm@0: 
Chris@23:     OutputDescriptor d2;
matthiasm@0:     d2.identifier = "tunedlogfreqspec";
matthiasm@0:     d2.name = "Tuned Log-Frequency Spectrum";
matthiasm@0:     d2.description = "A Log-Frequency Spectrum (constant Q) that is obtained by cosine filter mapping, then its tuned using the estimated tuning frequency.";
matthiasm@0:     d2.unit = "";
matthiasm@0:     d2.hasFixedBinCount = true;
matthiasm@0:     d2.binCount = 256;
matthiasm@0:     d2.hasKnownExtents = false;
matthiasm@0:     d2.isQuantized = false;
matthiasm@0:     d2.sampleType = OutputDescriptor::FixedSampleRate;
matthiasm@0:     d2.hasDuration = false;
matthiasm@0:     d2.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@0:     list.push_back(d2);
matthiasm@0:     
matthiasm@0:     OutputDescriptor d3;
matthiasm@0:     d3.identifier = "semitonespectrum";
matthiasm@0:     d3.name = "Semitone Spectrum";
matthiasm@0:     d3.description = "A semitone-spaced log-frequency spectrum derived from the third-of-a-semitone-spaced tuned log-frequency spectrum.";
matthiasm@0:     d3.unit = "";
matthiasm@0:     d3.hasFixedBinCount = true;
matthiasm@0:     d3.binCount = 84;
matthiasm@0:     d3.hasKnownExtents = false;
matthiasm@0:     d3.isQuantized = false;
matthiasm@0:     d3.sampleType = OutputDescriptor::FixedSampleRate;
matthiasm@0:     d3.hasDuration = false;
matthiasm@0:     d3.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@0:     list.push_back(d3);
matthiasm@0:     
matthiasm@0:     OutputDescriptor d4;
matthiasm@0:     d4.identifier = "chroma";
matthiasm@0:     d4.name = "Chromagram";
matthiasm@0:     d4.description = "Tuning-adjusted chromagram from NNLS soft transcription, with an emphasis on the medium note range.";
matthiasm@0:     d4.unit = "";
matthiasm@0:     d4.hasFixedBinCount = true;
matthiasm@0:     d4.binCount = 12;
matthiasm@0:     d4.binNames = chromanames;
matthiasm@0:     d4.hasKnownExtents = false;
matthiasm@0:     d4.isQuantized = false;
matthiasm@0:     d4.sampleType = OutputDescriptor::FixedSampleRate;
matthiasm@0:     d4.hasDuration = false;
matthiasm@0:     d4.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@0:     list.push_back(d4);
matthiasm@0:     
matthiasm@0:     OutputDescriptor d5;
matthiasm@0:     d5.identifier = "basschroma";
matthiasm@0:     d5.name = "Bass Chromagram";
matthiasm@0:     d5.description = "Tuning-adjusted bass chromagram from NNLS soft transcription, with an emphasis on the bass note range.";
matthiasm@0:     d5.unit = "";
matthiasm@0:     d5.hasFixedBinCount = true;
matthiasm@0:     d5.binCount = 12;
matthiasm@0:     d5.binNames = chromanames;
matthiasm@0:     d5.hasKnownExtents = false;
matthiasm@0:     d5.isQuantized = false;
matthiasm@0:     d5.sampleType = OutputDescriptor::FixedSampleRate;
matthiasm@0:     d5.hasDuration = false;
matthiasm@0:     d5.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@0:     list.push_back(d5);
matthiasm@0:     
matthiasm@0:     OutputDescriptor d6;
matthiasm@0:     d6.identifier = "bothchroma";
matthiasm@0:     d6.name = "Chromagram and Bass Chromagram";
matthiasm@0:     d6.description = "Tuning-adjusted chromagram and bass chromagram (stacked on top of each other) from NNLS soft transcription.";
matthiasm@0:     d6.unit = "";
matthiasm@0:     d6.hasFixedBinCount = true;
matthiasm@0:     d6.binCount = 24;
matthiasm@0:     d6.binNames = bothchromanames;
matthiasm@0:     d6.hasKnownExtents = false;
matthiasm@0:     d6.isQuantized = false;
matthiasm@0:     d6.sampleType = OutputDescriptor::FixedSampleRate;
matthiasm@0:     d6.hasDuration = false;
matthiasm@0:     d6.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@0:     list.push_back(d6);
matthiasm@0:     
matthiasm@0:     OutputDescriptor d7;
matthiasm@0:     d7.identifier = "simplechord";
matthiasm@0:     d7.name = "Simple Chord Estimate";
matthiasm@0:     d7.description = "A simple chord estimate based on the inner product of chord templates with the smoothed chroma.";
matthiasm@0:     d7.unit = "";
matthiasm@0:     d7.hasFixedBinCount = true;
matthiasm@0:     d7.binCount = 0;
matthiasm@0:     d7.hasKnownExtents = false;
matthiasm@0:     d7.isQuantized = false;
matthiasm@0:     d7.sampleType = OutputDescriptor::VariableSampleRate;
matthiasm@0:     d7.hasDuration = false;
matthiasm@0:     d7.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@0:     list.push_back(d7);
matthiasm@0:     
Chris@23:     //     
Chris@23:     //     OutputDescriptor d9;
Chris@23:     //     d9.identifier = "inconsistencysegment";
Chris@23:     //     d9.name = "Harmonic inconsistency segmenter";
Chris@23:     //     d9.description = "Segments the audio based on the harmonic inconsistency value into speech and music.";
Chris@23:     //     d9.unit = "";
Chris@23:     //     d9.hasFixedBinCount = true;
Chris@23:     //     d9.binCount = 0;
Chris@23:     //     d9.hasKnownExtents = true;
Chris@23:     //     d9.minValue = 0.1;
Chris@23:     // d9.maxValue = 0.9;
Chris@23:     //     d9.isQuantized = false;
Chris@23:     //     d9.sampleType = OutputDescriptor::VariableSampleRate;
Chris@23:     //     d9.hasDuration = false;
Chris@23:     //     d9.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
Chris@23:     //     list.push_back(d9);
Chris@23:     // 
Chris@23:     OutputDescriptor d10;
Chris@23:     d10.identifier = "localtuning";
Chris@23:     d10.name = "Local tuning";
Chris@23:     d10.description = "Tuning based on the history up to this timestamp.";
Chris@23:     d10.unit = "Hz";
Chris@23:     d10.hasFixedBinCount = true;
Chris@23:     d10.binCount = 1;
Chris@23:     d10.hasKnownExtents = true;
Chris@23:     d10.minValue = 427.47;
Chris@23:     d10.maxValue = 452.89;
Chris@23:     d10.isQuantized = false;
Chris@23:     d10.sampleType = OutputDescriptor::FixedSampleRate;
Chris@23:     d10.hasDuration = false;
Chris@23:     // d10.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
Chris@23:     list.push_back(d10);
matthiasm@17: 
Chris@23:     OutputDescriptor d8;
matthiasm@17:     d8.identifier = "harmonicchange";
matthiasm@17:     d8.name = "Harmonic change value";
matthiasm@17:     d8.description = "Harmonic change.";
matthiasm@17:     d8.unit = "";
matthiasm@17:     d8.hasFixedBinCount = true;
matthiasm@17:     d8.binCount = 1;
matthiasm@17:     d8.hasKnownExtents = true;
Chris@23:     d8.minValue = 0.0;
Chris@23:     d8.maxValue = 0.999;
matthiasm@17:     d8.isQuantized = false;
matthiasm@17:     d8.sampleType = OutputDescriptor::FixedSampleRate;
matthiasm@17:     d8.hasDuration = false;
matthiasm@17:     // d8.sampleRate = (m_stepSize == 0) ? m_inputSampleRate/2048 : m_inputSampleRate/m_stepSize;
matthiasm@17:     list.push_back(d8);
matthiasm@1:   
matthiasm@0:     return list;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: 
matthiasm@0: bool
matthiasm@0: NNLSChroma::initialise(size_t channels, size_t stepSize, size_t blockSize)
matthiasm@0: {	
Chris@23:     if (debug_on) {
Chris@23:         cerr << "--> initialise";
Chris@23:     }
matthiasm@1: 	
matthiasm@0:     if (channels < getMinChannelCount() ||
matthiasm@0: 	channels > getMaxChannelCount()) return false;
matthiasm@0:     m_blockSize = blockSize;
matthiasm@0:     m_stepSize = stepSize;
matthiasm@0:     frameCount = 0;
Chris@23:     int tempn = 256 * m_blockSize/2;
Chris@23:     // cerr << "length of tempkernel : " <<  tempn << endl;
Chris@23:     float *tempkernel;
matthiasm@1: 
Chris@23:     tempkernel = new float[tempn];
matthiasm@1: 
Chris@23:     logFreqMatrix(m_inputSampleRate, m_blockSize, tempkernel);
Chris@23:     m_kernelValue.clear();
Chris@23:     m_kernelFftIndex.clear();
Chris@23:     m_kernelNoteIndex.clear();
Chris@23:     int countNonzero = 0;
Chris@23:     for (unsigned iNote = 0; iNote < nNote; ++iNote) { // I don't know if this is wise: manually making a sparse matrix
Chris@23:         for (unsigned iFFT = 0; iFFT < blockSize/2; ++iFFT) {
Chris@23:             if (tempkernel[iFFT + blockSize/2 * iNote] > 0) {
Chris@23:                 m_kernelValue.push_back(tempkernel[iFFT + blockSize/2 * iNote]);
Chris@23:                 if (tempkernel[iFFT + blockSize/2 * iNote] > 0) {
Chris@23:                     countNonzero++;
Chris@23:                 }
Chris@23:                 m_kernelFftIndex.push_back(iFFT);
Chris@23:                 m_kernelNoteIndex.push_back(iNote);				
Chris@23:             }
Chris@23:         }
Chris@23:     }
Chris@23:     // cerr << "nonzero count : " << countNonzero << endl;
Chris@23:     delete [] tempkernel;
Chris@23:     ofstream myfile;
Chris@23:     myfile.open ("matrix.txt");
matthiasm@3:     // myfile << "Writing this to a file.\n";	
Chris@23:     for (int i = 0; i < nNote * 84; ++i) {
Chris@23:         myfile << m_dict[i] << endl;		
Chris@23:     }
matthiasm@3:     myfile.close();
matthiasm@0:     return true;
matthiasm@0: }
matthiasm@0: 
matthiasm@0: void
matthiasm@0: NNLSChroma::reset()
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> reset";
matthiasm@4: 	
matthiasm@0:     // Clear buffers, reset stored values, etc
Chris@23:     frameCount = 0;
Chris@23:     m_dictID = 0;
Chris@23:     m_fl.clear();
Chris@23:     m_meanTuning0 = 0;
Chris@23:     m_meanTuning1 = 0;
Chris@23:     m_meanTuning2 = 0;
Chris@23:     m_localTuning0 = 0;
Chris@23:     m_localTuning1 = 0;
Chris@23:     m_localTuning2 = 0;
Chris@23:     m_localTuning.clear();
matthiasm@0: }
matthiasm@0: 
matthiasm@0: NNLSChroma::FeatureSet
matthiasm@0: NNLSChroma::process(const float *const *inputBuffers, Vamp::RealTime timestamp)
matthiasm@0: {   
Chris@23:     if (debug_on) cerr << "--> process" << endl;
Chris@23:     frameCount++;   
Chris@23:     float *magnitude = new float[m_blockSize/2];
matthiasm@0: 	
Chris@23:     Feature f10; // local tuning
Chris@23:     f10.hasTimestamp = true;
Chris@23:     f10.timestamp = timestamp;
Chris@23:     const float *fbuf = inputBuffers[0];	
Chris@23:     float energysum = 0;
Chris@23:     // make magnitude
Chris@23:     float maxmag = -10000;
Chris@23:     for (size_t iBin = 0; iBin < m_blockSize/2; iBin++) {
Chris@23:         magnitude[iBin] = sqrt(fbuf[2 * iBin] * fbuf[2 * iBin] + 
Chris@23:                                fbuf[2 * iBin + 1] * fbuf[2 * iBin + 1]);		
Chris@23:         if (maxmag < magnitude[iBin]) maxmag = magnitude[iBin];
Chris@23:         if (m_rollon > 0) {
Chris@23:             energysum += pow(magnitude[iBin],2);
Chris@23:         }
Chris@23:     }
matthiasm@14: 	
Chris@23:     float cumenergy = 0;
Chris@23:     if (m_rollon > 0) {
Chris@23:         for (size_t iBin = 2; iBin < m_blockSize/2; iBin++) {
Chris@23:             cumenergy +=  pow(magnitude[iBin],2);
Chris@23:             if (cumenergy < energysum * m_rollon) magnitude[iBin-2] = 0;
Chris@23:             else break;
Chris@23:         }
Chris@23:     }
matthiasm@17: 	
Chris@23:     if (maxmag < 2) {
Chris@23:         // cerr << "timestamp " << timestamp << ": very low magnitude, setting magnitude to all zeros" << endl;
Chris@23:         for (size_t iBin = 0; iBin < m_blockSize/2; iBin++) {
Chris@23:             magnitude[iBin] = 0;
Chris@23:         }
Chris@23:     }
matthiasm@4: 		
Chris@23:     // note magnitude mapping using pre-calculated matrix
Chris@23:     float *nm  = new float[nNote]; // note magnitude
Chris@23:     for (size_t iNote = 0; iNote < nNote; iNote++) {
Chris@23:         nm[iNote] = 0; // initialise as 0
Chris@23:     }
Chris@23:     int binCount = 0;
Chris@23:     for (vector<float>::iterator it = m_kernelValue.begin(); it != m_kernelValue.end(); ++it) {
Chris@23:         // cerr << ".";
Chris@23:         nm[m_kernelNoteIndex[binCount]] += magnitude[m_kernelFftIndex[binCount]] * m_kernelValue[binCount];
Chris@23:         // cerr << m_kernelFftIndex[binCount] << " -- " << magnitude[m_kernelFftIndex[binCount]] << " -- "<< m_kernelValue[binCount] << endl;
Chris@23:         binCount++;	
Chris@23:     }
Chris@23:     // cerr << nm[20];
Chris@23:     // cerr << endl;
matthiasm@0: 	
matthiasm@0: 	
matthiasm@0:     float one_over_N = 1.0/frameCount;
matthiasm@0:     // update means of complex tuning variables
matthiasm@0:     m_meanTuning0 *= float(frameCount-1)*one_over_N;
matthiasm@0:     m_meanTuning1 *= float(frameCount-1)*one_over_N;
matthiasm@0:     m_meanTuning2 *= float(frameCount-1)*one_over_N;
matthiasm@0: 	
matthiasm@0:     for (int iTone = 0; iTone < 160; iTone = iTone + 3) {
matthiasm@0:         m_meanTuning0 += nm[iTone + 0]*one_over_N;
matthiasm@0:     	m_meanTuning1 += nm[iTone + 1]*one_over_N;
matthiasm@0:     	m_meanTuning2 += nm[iTone + 2]*one_over_N;
Chris@23:         float ratioOld = 0.997;
matthiasm@3:         m_localTuning0 *= ratioOld; m_localTuning0 += nm[iTone + 0] * (1 - ratioOld);
matthiasm@3:         m_localTuning1 *= ratioOld; m_localTuning1 += nm[iTone + 1] * (1 - ratioOld);
matthiasm@3:         m_localTuning2 *= ratioOld; m_localTuning2 += nm[iTone + 2] * (1 - ratioOld);
matthiasm@0:     }
matthiasm@0: 	
matthiasm@0:     // if (m_tuneLocal) {
Chris@23:     // local tuning
Chris@23:     float localTuningImag = sinvalue * m_localTuning1 - sinvalue * m_localTuning2;
Chris@23:     float localTuningReal = m_localTuning0 + cosvalue * m_localTuning1 + cosvalue * m_localTuning2;
Chris@23:     float normalisedtuning = atan2(localTuningImag, localTuningReal)/(2*M_PI);
Chris@23:     m_localTuning.push_back(normalisedtuning);
Chris@23:     float tuning440 = 440 * pow(2,normalisedtuning/12);
Chris@23:     f10.values.push_back(tuning440);
Chris@23:     // cerr << tuning440 << endl;
matthiasm@0:     // }
matthiasm@0:     
Chris@23:     Feature f1; // logfreqspec
Chris@23:     f1.hasTimestamp = true;
matthiasm@0:     f1.timestamp = timestamp;
Chris@23:     for (size_t iNote = 0; iNote < nNote; iNote++) {
Chris@23:         f1.values.push_back(nm[iNote]);
Chris@23:     }
matthiasm@0: 	
Chris@23:     FeatureSet fs;
Chris@23:     fs[1].push_back(f1);
matthiasm@3:     fs[8].push_back(f10);
matthiasm@0: 
matthiasm@0:     // deletes
matthiasm@0:     delete[] magnitude;
matthiasm@0:     delete[] nm;
matthiasm@0: 
matthiasm@0:     m_fl.push_back(f1); // remember note magnitude for getRemainingFeatures
Chris@23:     char * pPath;
Chris@23:     pPath = getenv ("VAMP_PATH");	
matthiasm@7: 	
matthiasm@7: 	
Chris@23:     return fs;	
matthiasm@0: }
matthiasm@0: 
matthiasm@0: NNLSChroma::FeatureSet
matthiasm@0: NNLSChroma::getRemainingFeatures()
matthiasm@0: {
Chris@23:     if (debug_on) cerr << "--> getRemainingFeatures" << endl;
Chris@23:     FeatureSet fsOut;
Chris@23:     if (m_fl.size() == 0) return fsOut;
Chris@23:     int nChord = m_chordnames.size();
Chris@23:     // 
Chris@23:     /**  Calculate Tuning
Chris@23:          calculate tuning from (using the angle of the complex number defined by the 
Chris@23:          cumulative mean real and imag values)
Chris@23:     **/
Chris@23:     float meanTuningImag = sinvalue * m_meanTuning1 - sinvalue * m_meanTuning2;
Chris@23:     float meanTuningReal = m_meanTuning0 + cosvalue * m_meanTuning1 + cosvalue * m_meanTuning2;
Chris@23:     float cumulativetuning = 440 * pow(2,atan2(meanTuningImag, meanTuningReal)/(24*M_PI));
Chris@23:     float normalisedtuning = atan2(meanTuningImag, meanTuningReal)/(2*M_PI);
Chris@23:     int intShift = floor(normalisedtuning * 3);
Chris@23:     float intFactor = normalisedtuning * 3 - intShift; // intFactor is a really bad name for this
matthiasm@1: 		    
Chris@23:     char buffer0 [50];
matthiasm@1: 		
Chris@23:     sprintf(buffer0, "estimated tuning: %0.1f Hz", cumulativetuning);
matthiasm@1: 		    
Chris@23:     // cerr << "normalisedtuning: " << normalisedtuning << '\n';
matthiasm@1: 		    
Chris@23:     // push tuning to FeatureSet fsOut
Chris@23:     Feature f0; // tuning
Chris@23:     f0.hasTimestamp = true;
Chris@23:     f0.timestamp = Vamp::RealTime::frame2RealTime(0, lrintf(m_inputSampleRate));;
Chris@23:     f0.label = buffer0;
Chris@23:     fsOut[0].push_back(f0);  
matthiasm@1: 		    
Chris@23:     /** Tune Log-Frequency Spectrogram
Chris@23:         calculate a tuned log-frequency spectrogram (f2): use the tuning estimated above (kinda f0) to 
Chris@23:         perform linear interpolation on the existing log-frequency spectrogram (kinda f1).
Chris@23:     **/
Chris@23:     cerr << endl << "[NNLS Chroma Plugin] Tuning Log-Frequency Spectrogram ... ";
matthiasm@13: 					
Chris@23:     float tempValue = 0;
Chris@23:     float dbThreshold = 0; // relative to the background spectrum
Chris@23:     float thresh = pow(10,dbThreshold/20);
Chris@23:     // cerr << "tune local ? " << m_tuneLocal << endl;
Chris@23:     int count = 0;
matthiasm@1: 		
Chris@23:     for (FeatureList::iterator i = m_fl.begin(); i != m_fl.end(); ++i) {
Chris@23:         Feature f1 = *i;
Chris@23:         Feature f2; // tuned log-frequency spectrum
Chris@23:         f2.hasTimestamp = true;
Chris@23:         f2.timestamp = f1.timestamp;
Chris@23:         f2.values.push_back(0.0); f2.values.push_back(0.0); // set lower edge to zero
matthiasm@1: 		
Chris@23:         if (m_tuneLocal) {
Chris@23:             intShift = floor(m_localTuning[count] * 3);
Chris@23:             intFactor = m_localTuning[count] * 3 - intShift; // intFactor is a really bad name for this
Chris@23:         }
matthiasm@1: 		        
Chris@23:         // cerr << intShift << " " << intFactor << endl;
matthiasm@1: 		        
Chris@23:         for (unsigned k = 2; k < f1.values.size() - 3; ++k) { // interpolate all inner bins
Chris@23:             tempValue = f1.values[k + intShift] * (1-intFactor) + f1.values[k+intShift+1] * intFactor;
Chris@23:             f2.values.push_back(tempValue);
Chris@23:         }
matthiasm@1: 		        
Chris@23:         f2.values.push_back(0.0); f2.values.push_back(0.0); f2.values.push_back(0.0); // upper edge
Chris@23:         vector<float> runningmean = SpecialConvolution(f2.values,hw);
Chris@23:         vector<float> runningstd;
Chris@23:         for (int i = 0; i < 256; i++) { // first step: squared values into vector (variance)
Chris@23:             runningstd.push_back((f2.values[i] - runningmean[i]) * (f2.values[i] - runningmean[i]));
Chris@23:         }
Chris@23:         runningstd = SpecialConvolution(runningstd,hw); // second step convolve
Chris@23:         for (int i = 0; i < 256; i++) { 
Chris@23:             runningstd[i] = sqrt(runningstd[i]); // square root to finally have running std
Chris@23:             if (runningstd[i] > 0) {
Chris@23:                 // f2.values[i] = (f2.values[i] / runningmean[i]) > thresh ? 
Chris@23:                 // 		                    (f2.values[i] - runningmean[i]) / pow(runningstd[i],m_paling) : 0;
Chris@23:                 f2.values[i] = (f2.values[i] - runningmean[i]) > 0 ?
Chris@23:                     (f2.values[i] - runningmean[i]) / pow(runningstd[i],m_paling) : 0;
Chris@23:             }
Chris@23:             if (f2.values[i] < 0) {
Chris@23:                 cerr << "ERROR: negative value in logfreq spectrum" << endl;
Chris@23:             }
Chris@23:         }
Chris@23:         fsOut[2].push_back(f2);
Chris@23:         count++;
Chris@23:     }
Chris@23:     cerr << "done." << endl;
matthiasm@1: 	    
Chris@23:     /** Semitone spectrum and chromagrams
Chris@23:         Semitone-spaced log-frequency spectrum derived from the tuned log-freq spectrum above. the spectrum
Chris@23:         is inferred using a non-negative least squares algorithm.
Chris@23:         Three different kinds of chromagram are calculated, "treble", "bass", and "both" (which means 
Chris@23:         bass and treble stacked onto each other).
Chris@23:     **/
Chris@23:     if (m_dictID == 1) {
Chris@23:         cerr << "[NNLS Chroma Plugin] Mapping to semitone spectrum and chroma ... ";
Chris@23:     } else {
Chris@23:         cerr << "[NNLS Chroma Plugin] Performing NNLS and mapping to chroma ... ";
Chris@23:     }
matthiasm@13: 
matthiasm@1: 	    
Chris@23:     vector<vector<float> > chordogram;
Chris@23:     vector<vector<int> > scoreChordogram;
Chris@23:     vector<float> chordchange = vector<float>(fsOut[2].size(),0);
Chris@23:     vector<float> oldchroma = vector<float>(12,0);
Chris@23:     vector<float> oldbasschroma = vector<float>(12,0);
Chris@23:     count = 0;
matthiasm@9: 
Chris@23:     for (FeatureList::iterator it = fsOut[2].begin(); it != fsOut[2].end(); ++it) {
Chris@23:         Feature f2 = *it; // logfreq spectrum
Chris@23:         Feature f3; // semitone spectrum
Chris@23:         Feature f4; // treble chromagram
Chris@23:         Feature f5; // bass chromagram
Chris@23:         Feature f6; // treble and bass chromagram
matthiasm@1: 	
Chris@23:         f3.hasTimestamp = true;
Chris@23:         f3.timestamp = f2.timestamp;
matthiasm@1: 	        
Chris@23:         f4.hasTimestamp = true;
Chris@23:         f4.timestamp = f2.timestamp;
matthiasm@1: 	        
Chris@23:         f5.hasTimestamp = true;
Chris@23:         f5.timestamp = f2.timestamp;
matthiasm@1: 	        
Chris@23:         f6.hasTimestamp = true;
Chris@23:         f6.timestamp = f2.timestamp;
matthiasm@1: 	        
Chris@29: 			float b[256];
matthiasm@1: 	
Chris@23:         bool some_b_greater_zero = false;
Chris@23:         float sumb = 0;
Chris@23:         for (int i = 0; i < 256; i++) {
Chris@23:             // b[i] = m_dict[(256 * count + i) % (256 * 84)];
Chris@23:             b[i] = f2.values[i];
Chris@23:             sumb += b[i];
Chris@23:             if (b[i] > 0) {
Chris@23:                 some_b_greater_zero = true;
Chris@23:             }            
Chris@23:         }
matthiasm@1: 	    
Chris@23:         // here's where the non-negative least squares algorithm calculates the note activation x
matthiasm@1: 	
Chris@23:         vector<float> chroma = vector<float>(12, 0);
Chris@23:         vector<float> basschroma = vector<float>(12, 0);
Chris@23:         float currval;
Chris@23:         unsigned iSemitone = 0;
matthiasm@1: 			
Chris@23:         if (some_b_greater_zero) {
Chris@23:             if (m_dictID == 1) {
Chris@23:                 for (unsigned iNote = 2; iNote < nNote - 2; iNote += 3) {
Chris@23:                     currval = 0;
Chris@23:                     currval += b[iNote + 1 + -1] * 0.5;						
Chris@23:                     currval += b[iNote + 1 +  0] * 1.0;						
Chris@23:                     currval += b[iNote + 1 +  1] * 0.5;						
Chris@23:                     f3.values.push_back(currval);
Chris@23:                     chroma[iSemitone % 12] += currval * treblewindow[iSemitone];
Chris@23:                     basschroma[iSemitone % 12] += currval * basswindow[iSemitone];
Chris@23:                     iSemitone++;
Chris@23:                 }
matthiasm@1: 		        
Chris@23:             } else {
Chris@29: 					float x[84+1000];
Chris@23:                 for (int i = 1; i < 1084; ++i) x[i] = 1.0;
Chris@23:                 vector<int> signifIndex;
Chris@23:                 int index=0;
Chris@23:                 sumb /= 84.0;
Chris@23:                 for (unsigned iNote = 2; iNote < nNote - 2; iNote += 3) {
Chris@23:                     float currval = 0;
Chris@23:                     currval += b[iNote + 1 + -1];						
Chris@23:                     currval += b[iNote + 1 +  0];						
Chris@23:                     currval += b[iNote + 1 +  1];
Chris@23:                     if (currval > 0) signifIndex.push_back(index);
Chris@23:                     f3.values.push_back(0); // fill the values, change later
Chris@23:                     index++;
Chris@23:                 }
Chris@29: 				    float rnorm;
Chris@29: 				    float w[84+1000];
Chris@29: 				    float zz[84+1000];
Chris@23:                 int indx[84+1000];
Chris@23:                 int mode;
Chris@23:                 int dictsize = 256*signifIndex.size();
Chris@23:                 // cerr << "dictsize is " << dictsize << "and values size" << f3.values.size()<< endl;
Chris@29: 					float *curr_dict = new float[dictsize];
Chris@23:                 for (unsigned iNote = 0; iNote < signifIndex.size(); ++iNote) {
Chris@23:                     for (unsigned iBin = 0; iBin < 256; iBin++) {
Chris@23:                         curr_dict[iNote * 256 + iBin] = 1.0 * m_dict[signifIndex[iNote] * 256 + iBin];
Chris@23:                     }
Chris@23:                 }
Chris@29: 					nnls(curr_dict, nNote, nNote, signifIndex.size(), b, x, &rnorm, w, zz, indx, &mode);
Chris@23:                 delete [] curr_dict;
Chris@23:                 for (unsigned iNote = 0; iNote < signifIndex.size(); ++iNote) {
Chris@23:                     f3.values[signifIndex[iNote]] = x[iNote];
Chris@23:                     // cerr << mode << endl;
Chris@23:                     chroma[signifIndex[iNote] % 12] += x[iNote] * treblewindow[signifIndex[iNote]];
Chris@23:                     basschroma[signifIndex[iNote] % 12] += x[iNote] * basswindow[signifIndex[iNote]];
Chris@23:                 }
Chris@23:             }	
Chris@23:         }
matthiasm@13: 			
matthiasm@10:             
matthiasm@12: 	        
matthiasm@13: 			
Chris@23:         f4.values = chroma; 
Chris@23:         f5.values = basschroma;
Chris@23:         chroma.insert(chroma.begin(), basschroma.begin(), basschroma.end()); // just stack the both chromas 
Chris@23:         f6.values = chroma; 
matthiasm@1: 	        
Chris@23:         if (m_doNormalizeChroma > 0) {
Chris@23:             vector<float> chromanorm = vector<float>(3,0);			
Chris@23:             switch (int(m_doNormalizeChroma)) {
Chris@23:             case 0: // should never end up here
Chris@23:                 break;
Chris@23:             case 1:
Chris@23:                 chromanorm[0] = *max_element(f4.values.begin(), f4.values.end());
Chris@23:                 chromanorm[1] = *max_element(f5.values.begin(), f5.values.end());
Chris@23:                 chromanorm[2] = max(chromanorm[0], chromanorm[1]);
Chris@23:                 break;
Chris@23:             case 2:
Chris@23:                 for (vector<float>::iterator it = f4.values.begin(); it != f4.values.end(); ++it) {
Chris@23:                     chromanorm[0] += *it; 						
Chris@23:                 }
Chris@23:                 for (vector<float>::iterator it = f5.values.begin(); it != f5.values.end(); ++it) {
Chris@23:                     chromanorm[1] += *it; 						
Chris@23:                 }
Chris@23:                 for (vector<float>::iterator it = f6.values.begin(); it != f6.values.end(); ++it) {
Chris@23:                     chromanorm[2] += *it; 						
Chris@23:                 }
Chris@23:                 break;
Chris@23:             case 3:
Chris@23:                 for (vector<float>::iterator it = f4.values.begin(); it != f4.values.end(); ++it) {
Chris@23:                     chromanorm[0] += pow(*it,2); 						
Chris@23:                 }
Chris@23:                 chromanorm[0] = sqrt(chromanorm[0]);
Chris@23:                 for (vector<float>::iterator it = f5.values.begin(); it != f5.values.end(); ++it) {
Chris@23:                     chromanorm[1] += pow(*it,2); 						
Chris@23:                 }
Chris@23:                 chromanorm[1] = sqrt(chromanorm[1]);
Chris@23:                 for (vector<float>::iterator it = f6.values.begin(); it != f6.values.end(); ++it) {
Chris@23:                     chromanorm[2] += pow(*it,2); 						
Chris@23:                 }
Chris@23:                 chromanorm[2] = sqrt(chromanorm[2]);
Chris@23:                 break;
Chris@23:             }
Chris@23:             if (chromanorm[0] > 0) {
Chris@23:                 for (int i = 0; i < f4.values.size(); i++) {
Chris@23:                     f4.values[i] /= chromanorm[0];
Chris@23:                 }
Chris@23:             }
Chris@23:             if (chromanorm[1] > 0) {
Chris@23:                 for (int i = 0; i < f5.values.size(); i++) {
Chris@23:                     f5.values[i] /= chromanorm[1];
Chris@23:                 }
Chris@23:             }
Chris@23:             if (chromanorm[2] > 0) {
Chris@23:                 for (int i = 0; i < f6.values.size(); i++) {
Chris@23:                     f6.values[i] /= chromanorm[2];
Chris@23:                 }
Chris@23:             }
matthiasm@13: 				
Chris@23:         }
matthiasm@13: 	
Chris@23:         // local chord estimation
Chris@23:         vector<float> currentChordSalience;
Chris@23:         float tempchordvalue = 0;
Chris@23:         float sumchordvalue = 0;
matthiasm@9: 	        
Chris@23:         for (int iChord = 0; iChord < nChord; iChord++) {
Chris@23:             tempchordvalue = 0;
Chris@23:             for (int iBin = 0; iBin < 12; iBin++) {
Chris@23:                 tempchordvalue += m_chorddict[24 * iChord + iBin] * chroma[iBin];
Chris@23:             }
Chris@23:             for (int iBin = 12; iBin < 24; iBin++) {
Chris@23:                 tempchordvalue += m_chorddict[24 * iChord + iBin] * chroma[iBin];
Chris@23:             }
Chris@23:             sumchordvalue+=tempchordvalue;
Chris@23:             currentChordSalience.push_back(tempchordvalue);
Chris@23:         }
Chris@23:         if (sumchordvalue > 0) {
Chris@23:             for (int iChord = 0; iChord < nChord; iChord++) {
Chris@23:                 currentChordSalience[iChord] /= sumchordvalue;
Chris@23:             }
Chris@23:         } else {
Chris@23:             currentChordSalience[nChord-1] = 1.0;
Chris@23:         }
Chris@23:         chordogram.push_back(currentChordSalience);
matthiasm@1: 	        
Chris@23:         fsOut[3].push_back(f3);
Chris@23:         fsOut[4].push_back(f4);
Chris@23:         fsOut[5].push_back(f5);
Chris@23:         fsOut[6].push_back(f6);
Chris@23:         count++;
Chris@23:     }
Chris@23:     cerr << "done." << endl;
matthiasm@13: 		
matthiasm@10: 
Chris@23:     /* Simple chord estimation
Chris@23:        I just take the local chord estimates ("currentChordSalience") and average them over time, then
Chris@23:        take the maximum. Very simple, don't do this at home...
Chris@23:     */
Chris@23:     cerr << "[NNLS Chroma Plugin] Chord Estimation ... ";
Chris@23:     count = 0; 
Chris@23:     int halfwindowlength = m_inputSampleRate / m_stepSize;
Chris@23:     vector<int> chordSequence;
Chris@23:     for (FeatureList::iterator it = fsOut[6].begin(); it != fsOut[6].end(); ++it) { // initialise the score chordogram
Chris@23:         vector<int> temp = vector<int>(nChord,0);
Chris@23:         scoreChordogram.push_back(temp);
Chris@23:     }
Chris@23:     for (FeatureList::iterator it = fsOut[6].begin(); it < fsOut[6].end()-2*halfwindowlength-1; ++it) {		
Chris@23:         int startIndex = count + 1;
Chris@23:         int endIndex = count + 2 * halfwindowlength;
matthiasm@10: 			
Chris@23:         float chordThreshold = 2.5/nChord;//*(2*halfwindowlength+1);
matthiasm@10:             
Chris@23:         vector<int> chordCandidates;
Chris@23:         for (unsigned iChord = 0; iChord < nChord-1; iChord++) {
Chris@23:             // float currsum = 0;
Chris@23:             // for (unsigned iFrame = startIndex; iFrame < endIndex; ++iFrame) {
Chris@23:             //  currsum += chordogram[iFrame][iChord];
Chris@23:             // }
Chris@23:             //                 if (currsum > chordThreshold) chordCandidates.push_back(iChord);
Chris@23:             for (unsigned iFrame = startIndex; iFrame < endIndex; ++iFrame) {
Chris@23:                 if (chordogram[iFrame][iChord] > chordThreshold) {
Chris@23:                     chordCandidates.push_back(iChord);
Chris@23:                     break;
Chris@23:                 }                    
Chris@23:             }
Chris@23:         }
Chris@23:         chordCandidates.push_back(nChord-1);
Chris@23:         // cerr << chordCandidates.size() << endl;          
Chris@23: 	        
Chris@23:         float maxval = 0; // will be the value of the most salient *chord change* in this frame
Chris@23:         float maxindex = 0; //... and the index thereof
Chris@23:         unsigned bestchordL = nChord-1; // index of the best "left" chord
Chris@23:         unsigned bestchordR = nChord-1; // index of the best "right" chord
Chris@23:  	 		
Chris@23:         for (int iWF = 1; iWF < 2*halfwindowlength; ++iWF) {
Chris@23:             // now find the max values on both sides of iWF
Chris@23:             // left side:
Chris@23:             float maxL = 0;
Chris@23:             unsigned maxindL = nChord-1;
Chris@23:             for (unsigned kChord = 0; kChord < chordCandidates.size(); kChord++) {
Chris@23:                 unsigned iChord = chordCandidates[kChord];
Chris@23:                 float currsum = 0;
Chris@23:                 for (unsigned iFrame = 0; iFrame < iWF-1; ++iFrame) {
Chris@23:                     currsum += chordogram[count+iFrame][iChord];
matthiasm@10:                 }
Chris@23:                 if (iChord == nChord-1) currsum *= 0.8;
Chris@23:                 if (currsum > maxL) {
Chris@23:                     maxL = currsum;
Chris@23:                     maxindL = iChord;
Chris@23:                 }
Chris@23:             }				
Chris@23:             // right side:
Chris@23:             float maxR = 0;
Chris@23:             unsigned maxindR = nChord-1;
Chris@23:             for (unsigned kChord = 0; kChord < chordCandidates.size(); kChord++) {
Chris@23:                 unsigned iChord = chordCandidates[kChord];
Chris@23:                 float currsum = 0;
Chris@23:                 for (unsigned iFrame = iWF-1; iFrame < 2*halfwindowlength; ++iFrame) {
Chris@23:                     currsum += chordogram[count+iFrame][iChord];
Chris@23:                 }
Chris@23:                 if (iChord == nChord-1) currsum *= 0.8;
Chris@23:                 if (currsum > maxR) {
Chris@23:                     maxR = currsum;
Chris@23:                     maxindR = iChord;
Chris@23:                 }
Chris@23:             }
Chris@23:             if (maxL+maxR > maxval) {					
Chris@23:                 maxval = maxL+maxR;
Chris@23:                 maxindex = iWF;
Chris@23:                 bestchordL = maxindL;
Chris@23:                 bestchordR = maxindR;
Chris@23:             }
matthiasm@3: 				
Chris@23:         }
Chris@23:         // cerr << "maxindex: " << maxindex << ", bestchordR is " << bestchordR << ", of frame " << count << endl;
Chris@23:         // add a score to every chord-frame-point that was part of a maximum 
Chris@23:         for (unsigned iFrame = 0; iFrame < maxindex-1; ++iFrame) {
Chris@23:             scoreChordogram[iFrame+count][bestchordL]++;
Chris@23:         }
Chris@23:         for (unsigned iFrame = maxindex-1; iFrame < 2*halfwindowlength; ++iFrame) {
Chris@23:             scoreChordogram[iFrame+count][bestchordR]++;
Chris@23:         }
Chris@23:         if (bestchordL != bestchordR) chordchange[maxindex+count] += (halfwindowlength - abs(maxindex-halfwindowlength)) * 2.0 / halfwindowlength;
Chris@23:         count++;	
Chris@23:     }
Chris@23:     // cerr << "*******  agent finished   *******" << endl;
Chris@23:     count = 0;
Chris@23:     for (FeatureList::iterator it = fsOut[6].begin(); it != fsOut[6].end(); ++it) { 
Chris@23:         float maxval = 0; // will be the value of the most salient chord in this frame
Chris@23:         float maxindex = 0; //... and the index thereof
Chris@23:         for (unsigned iChord = 0; iChord < nChord; iChord++) {
Chris@23:             if (scoreChordogram[count][iChord] > maxval) {
Chris@23:                 maxval = scoreChordogram[count][iChord];
Chris@23:                 maxindex = iChord;
Chris@23:                 // cerr << iChord << endl;
Chris@23:             }
Chris@23:         }
Chris@23:         chordSequence.push_back(maxindex);
Chris@23:         // cerr << "before modefilter, maxindex: " << maxindex << endl;
Chris@23:         count++;
Chris@23:     }
Chris@23:     // cerr << "*******  mode filter done *******" << endl;
matthiasm@10: 
matthiasm@3: 	
Chris@23:     // mode filter on chordSequence
Chris@23:     count = 0;
Chris@23:     string oldChord = "";
Chris@23:     for (FeatureList::iterator it = fsOut[6].begin(); it != fsOut[6].end(); ++it) {
Chris@23:         Feature f6 = *it;
Chris@23:         Feature f7; // chord estimate
Chris@23:         f7.hasTimestamp = true;
Chris@23:         f7.timestamp = f6.timestamp;
Chris@23:         Feature f8; // chord estimate
Chris@23:         f8.hasTimestamp = true;
Chris@23:         f8.timestamp = f6.timestamp;
matthiasm@17: 			
Chris@23:         vector<int> chordCount = vector<int>(nChord,0);
Chris@23:         int maxChordCount = 0;
Chris@23:         int maxChordIndex = nChord-1;
Chris@23:         string maxChord;
Chris@23:         int startIndex = max(count - halfwindowlength/2,0);
Chris@23:         int endIndex = min(int(chordogram.size()), count + halfwindowlength/2);
Chris@23:         for (int i = startIndex; i < endIndex; i++) {				
Chris@23:             chordCount[chordSequence[i]]++;
Chris@23:             if (chordCount[chordSequence[i]] > maxChordCount) {
Chris@23:                 // cerr << "start index " << startIndex << endl;
Chris@23:                 maxChordCount++;
Chris@23:                 maxChordIndex = chordSequence[i];
Chris@23:                 maxChord = m_chordnames[maxChordIndex];
Chris@23:             }
Chris@23:         }
Chris@23:         // chordSequence[count] = maxChordIndex;
Chris@23:         // cerr << maxChordIndex << endl;
Chris@23:         f8.values.push_back(chordchange[count]/(halfwindowlength*2));
Chris@23:         // cerr << chordchange[count] << endl;
Chris@23:         fsOut[9].push_back(f8);
Chris@23:         if (oldChord != maxChord) {
Chris@23:             oldChord = maxChord;
matthiasm@3: 	
Chris@23:             // char buffer1 [50];
Chris@23:             // if (maxChordIndex < nChord - 1) {
Chris@23:             //     sprintf(buffer1, "%s%s", notenames[maxChordIndex % 12 + 12], chordtypes[maxChordIndex]);
Chris@23:             // } else {
Chris@23:             //     sprintf(buffer1, "N");
Chris@23:             // }
Chris@23:             // f7.label = buffer1;
Chris@23:             f7.label = m_chordnames[maxChordIndex];
Chris@23:             fsOut[7].push_back(f7);
Chris@23:         }
Chris@23:         count++;
Chris@23:     }
Chris@23:     Feature f7; // last chord estimate
Chris@23:     f7.hasTimestamp = true;
Chris@23:     f7.timestamp = fsOut[6][fsOut[6].size()-1].timestamp;
Chris@23:     f7.label = "N";
Chris@23:     fsOut[7].push_back(f7);
Chris@23:     cerr << "done." << endl;
Chris@23:     //     // musicity
Chris@23:     //     count = 0;
Chris@23:     //     int oldlabeltype = 0; // start value is 0, music is 1, speech is 2
Chris@23:     //     vector<float> musicityValue; 
Chris@23:     //     for (FeatureList::iterator it = fsOut[4].begin(); it != fsOut[4].end(); ++it) {
Chris@23:     //         Feature f4 = *it;
Chris@23:     //         
Chris@23:     //         int startIndex = max(count - musicitykernelwidth/2,0);
Chris@23:     //         int endIndex = min(int(chordogram.size()), startIndex + musicitykernelwidth - 1);
Chris@23:     //         float chromasum = 0;
Chris@23:     //         float diffsum = 0;
Chris@23:     //         for (int k = 0; k < 12; k++) {
Chris@23:     //             for (int i = startIndex + 1; i < endIndex; i++) {
Chris@23:     //                 chromasum += pow(fsOut[4][i].values[k],2);
Chris@23:     //                 diffsum += abs(fsOut[4][i-1].values[k] - fsOut[4][i].values[k]);
Chris@23:     //             }
Chris@23:     //         }
Chris@23:     //         diffsum /= chromasum;
Chris@23:     //         musicityValue.push_back(diffsum);        
Chris@23:     //         count++;
Chris@23:     //     }
Chris@23:     //     
Chris@23:     //     float musicityThreshold = 0.44;
Chris@23:     //     if (m_stepSize == 4096) {
Chris@23:     //         musicityThreshold = 0.74;
Chris@23:     //     }
Chris@23:     //     if (m_stepSize == 4410) {
Chris@23:     //         musicityThreshold = 0.77;
Chris@23:     //     }
Chris@23:     //     
Chris@23:     //     count = 0;
Chris@23:     //     for (FeatureList::iterator it = fsOut[4].begin(); it != fsOut[4].end(); ++it) {
Chris@23:     //         Feature f4 = *it;
Chris@23:     //         Feature f8; // musicity
Chris@23:     //         Feature f9; // musicity segmenter
Chris@23:     //         
Chris@23:     //         f8.hasTimestamp = true;
Chris@23:     //         f8.timestamp = f4.timestamp;
Chris@23:     //         f9.hasTimestamp = true;
Chris@23:     //         f9.timestamp = f4.timestamp;    
Chris@23:     //         
Chris@23:     //         int startIndex = max(count - musicitykernelwidth/2,0);
Chris@23:     //         int endIndex = min(int(chordogram.size()), startIndex + musicitykernelwidth - 1);
Chris@23:     //         int musicityCount = 0;
Chris@23:     //         for (int i = startIndex; i <= endIndex; i++) {
Chris@23:     //             if (musicityValue[i] > musicityThreshold) musicityCount++;
Chris@23:     //         }
Chris@23:     //         bool isSpeech = (2 * musicityCount > endIndex - startIndex + 1); 
Chris@23:     //         
Chris@23:     //         if (isSpeech) {
Chris@23:     //             if (oldlabeltype != 2) {
Chris@23:     //                 f9.label = "Speech";
Chris@23:     //                 fsOut[9].push_back(f9);
Chris@23:     //                 oldlabeltype = 2;
Chris@23:     //             }
Chris@23:     //         } else {
Chris@23:     //             if (oldlabeltype != 1) {
Chris@23:     //                 f9.label = "Music";
Chris@23:     //                 fsOut[9].push_back(f9);
Chris@23:     //                 oldlabeltype = 1;
Chris@23:     //             }
Chris@23:     //         }
Chris@23:     //         f8.values.push_back(musicityValue[count]);
Chris@23:     //         fsOut[8].push_back(f8);
Chris@23:     //         count++;
Chris@23:     //      }
Chris@23:     return fsOut;     
matthiasm@0: 
matthiasm@0: }
matthiasm@0: