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 plugin using the MATCH audio alignment
cannam@0:     algorithm.
cannam@0: 
cannam@0:     Centre for Digital Music, Queen Mary, University of London.
Chris@236:     Copyright (c) 2007-2020 Simon Dixon, Chris Cannam, and Queen Mary
Chris@230:     University of London, Copyright (c) 2014-2015 Tido GmbH.
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: #include "MatchVampPlugin.h"
cannam@0: 
Chris@74: #include "FeatureExtractor.h"
cannam@0: 
cannam@0: #include <vamp/vamp.h>
cannam@0: #include <vamp-sdk/RealTime.h>
cannam@0: 
cannam@0: #include <vector>
cannam@0: #include <algorithm>
cannam@0: 
Chris@235: using std::string;
Chris@235: 
cannam@0: //static int extant = 0;
cannam@0: 
cannam@0: #ifdef _WIN32
cannam@0: HANDLE
cannam@0: MatchVampPlugin::m_serialisingMutex;
cannam@0: #else
cannam@0: pthread_mutex_t 
cannam@0: MatchVampPlugin::m_serialisingMutex;
cannam@0: #endif
cannam@0: 
cannam@0: bool
cannam@0: MatchVampPlugin::m_serialisingMutexInitialised = false;
cannam@0: 
Chris@10: // We want to ensure our freq map / crossover bin in Matcher.cpp are
Chris@10: // always valid with a fixed FFT length in seconds, so must reject low
Chris@10: // sample rates
Chris@10: static float sampleRateMin = 5000.f;
Chris@10: 
Chris@52: static float defaultStepTime = 0.020f;
Chris@15: 
cannam@0: MatchVampPlugin::MatchVampPlugin(float inputSampleRate) :
cannam@0:     Plugin(inputSampleRate),
Chris@52:     m_stepSize(int(inputSampleRate * defaultStepTime + 0.001)),
Chris@15:     m_stepTime(defaultStepTime),
Chris@16:     m_blockSize(2048),
cannam@0:     m_serialise(false),
cannam@0:     m_begin(true),
Chris@17:     m_locked(false),
Chris@138:     m_smooth(false),
Chris@74:     m_frameNo(0),
Chris@113:     m_params(defaultStepTime),
Chris@113:     m_defaultParams(defaultStepTime),
Chris@216:     m_feParams(inputSampleRate),
Chris@223:     m_defaultFeParams(44100), // parameter descriptors can't depend on samplerate
Chris@243:     m_secondReferenceFrequency(m_defaultFeParams.referenceFrequency), // must be declared/initialised after m_defaultFeParams
Chris@103:     m_fcParams(),
Chris@143:     m_defaultFcParams(),
Chris@143:     m_dParams(),
Chris@143:     m_defaultDParams()
cannam@0: {
Chris@10:     if (inputSampleRate < sampleRateMin) {
Chris@10:         std::cerr << "MatchVampPlugin::MatchVampPlugin: input sample rate "
Chris@10:                   << inputSampleRate << " < min supported rate "
Chris@10:                   << sampleRateMin << ", plugin will refuse to initialise"
Chris@10:                   << std::endl;
Chris@10:     }
Chris@10: 
cannam@0:     if (!m_serialisingMutexInitialised) {
cannam@0:         m_serialisingMutexInitialised = true;
cannam@0: #ifdef _WIN32
cannam@0:         m_serialisingMutex = CreateMutex(NULL, FALSE, NULL);
cannam@0: #else
cannam@0:         pthread_mutex_init(&m_serialisingMutex, 0);
cannam@0: #endif
cannam@0:     }
cannam@0: 
Chris@107:     m_pipeline = 0;
cannam@0: //    std::cerr << "MatchVampPlugin::MatchVampPlugin(" << this << "): extant = " << ++extant << std::endl;
cannam@0: }
cannam@0: 
cannam@0: MatchVampPlugin::~MatchVampPlugin()
cannam@0: {
cannam@0: //    std::cerr << "MatchVampPlugin::~MatchVampPlugin(" << this << "): extant = " << --extant << std::endl;
cannam@0: 
Chris@107:     delete m_pipeline;
cannam@0: 
cannam@0:     if (m_locked) {
cannam@0: #ifdef _WIN32
cannam@0:         ReleaseMutex(m_serialisingMutex);
cannam@0: #else
cannam@0:         pthread_mutex_unlock(&m_serialisingMutex);
cannam@0: #endif
cannam@0:         m_locked = false;
cannam@0:     }
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: MatchVampPlugin::getIdentifier() const
cannam@0: {
cannam@0:     return "match";
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: MatchVampPlugin::getName() const
cannam@0: {
cannam@0:     return "Match Performance Aligner";
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: MatchVampPlugin::getDescription() const
cannam@0: {
cannam@0:     return "Calculate alignment between two performances in separate channel inputs";
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: MatchVampPlugin::getMaker() const
cannam@0: {
cannam@0:     return "Simon Dixon (plugin by Chris Cannam)";
cannam@0: }
cannam@0: 
cannam@0: int
cannam@0: MatchVampPlugin::getPluginVersion() const
cannam@0: {
Chris@137:     return 3;
cannam@0: }
cannam@0: 
cannam@0: string
cannam@0: MatchVampPlugin::getCopyright() const
cannam@0: {
cannam@0:     return "GPL";
cannam@0: }
cannam@0: 
cannam@0: MatchVampPlugin::ParameterList
cannam@0: MatchVampPlugin::getParameterDescriptors() const
cannam@0: {
cannam@0:     ParameterList list;
cannam@0: 
cannam@0:     ParameterDescriptor desc;
Chris@18: 
Chris@162:     desc.identifier = "freq1";
Chris@162:     desc.name = "Tuning frequency of first input";
Chris@241:     desc.description = "Tuning frequency (concert A) for the reference audio";
Chris@162:     desc.minValue = 220.0;
Chris@162:     desc.maxValue = 880.0;
Chris@188:     desc.defaultValue = float(m_defaultFeParams.referenceFrequency);
Chris@162:     desc.isQuantized = false;
Chris@162:     desc.unit = "Hz";
Chris@162:     list.push_back(desc);
Chris@162: 
Chris@162:     desc.identifier = "freq2";
Chris@162:     desc.name = "Tuning frequency of second input";
Chris@241:     desc.description = "Tuning frequency (concert A) for the other audio";
Chris@162:     desc.minValue = 220.0;
Chris@162:     desc.maxValue = 880.0;
Chris@188:     desc.defaultValue = float(m_defaultFeParams.referenceFrequency);
Chris@162:     desc.isQuantized = false;
Chris@162:     desc.unit = "Hz";
Chris@162:     list.push_back(desc);
Chris@162: 
Chris@177:     desc.identifier = "minfreq";
Chris@177:     desc.name = "Minimum frequency";
Chris@241:     desc.description = "Minimum frequency to include in features";
Chris@177:     desc.minValue = 0.0;
Chris@188:     desc.maxValue = float(m_inputSampleRate / 4.f);
Chris@188:     desc.defaultValue = float(m_defaultFeParams.minFrequency);
Chris@177:     desc.isQuantized = false;
Chris@177:     desc.unit = "Hz";
Chris@177:     list.push_back(desc);
Chris@177: 
Chris@177:     desc.identifier = "maxfreq";
Chris@177:     desc.name = "Maximum frequency";
Chris@241:     desc.description = "Maximum frequency to include in features";
Chris@177:     desc.minValue = 1000.0;
Chris@188:     desc.maxValue = float(m_inputSampleRate / 2.f);
Chris@188:     desc.defaultValue = float(m_defaultFeParams.maxFrequency);
Chris@177:     desc.isQuantized = false;
Chris@177:     desc.unit = "Hz";
Chris@177:     list.push_back(desc);
Chris@177:     
Chris@162:     desc.unit = "";
Chris@162:     
Chris@162:     desc.identifier = "usechroma";
Chris@162:     desc.name = "Feature type";
Chris@162:     desc.description = "Whether to use warped spectrogram or chroma frequency map";
cannam@0:     desc.minValue = 0;
cannam@0:     desc.maxValue = 1;
Chris@162:     desc.defaultValue = m_defaultFeParams.useChromaFrequencyMap ? 1 : 0;
Chris@162:     desc.isQuantized = true;
Chris@162:     desc.quantizeStep = 1;
Chris@162:     desc.valueNames.clear();
Chris@162:     desc.valueNames.push_back("Spectral");
Chris@162:     desc.valueNames.push_back("Chroma");
Chris@162:     list.push_back(desc);
Chris@162: 
Chris@162:     desc.valueNames.clear();
Chris@162: 
Chris@162:     desc.identifier = "usespecdiff";
Chris@162:     desc.name = "Use feature difference";
Chris@162:     desc.description = "Whether to use half-wave rectified feature-to-feature difference instead of straight spectral or chroma feature";
Chris@162:     desc.minValue = 0;
Chris@162:     desc.maxValue = 1;
Chris@188:     desc.defaultValue = float(m_defaultFcParams.order);
cannam@0:     desc.isQuantized = true;
cannam@0:     desc.quantizeStep = 1;
cannam@0:     list.push_back(desc);
cannam@0: 
Chris@18:     desc.identifier = "framenorm";
Chris@162:     desc.name = "Frame normalisation";
Chris@162:     desc.description = "Type of normalisation to use for features";
Chris@18:     desc.minValue = 0;
Chris@18:     desc.maxValue = 2;
Chris@188:     desc.defaultValue = float(m_defaultFcParams.norm);
Chris@18:     desc.isQuantized = true;
Chris@18:     desc.quantizeStep = 1;
Chris@18:     desc.valueNames.clear();
Chris@18:     desc.valueNames.push_back("None");
Chris@162:     desc.valueNames.push_back("Sum to 1");
Chris@162:     desc.valueNames.push_back("Long-term average");
Chris@18:     list.push_back(desc);
Chris@18:     desc.valueNames.clear();
Chris@188:     desc.defaultValue = float(m_defaultFcParams.silenceThreshold);
Chris@151: 
Chris@156:     desc.identifier = "metric";
Chris@156:     desc.name = "Distance metric";
Chris@241:     desc.description = "Metric for distance calculations";
Chris@156:     desc.minValue = 0;
Chris@157:     desc.maxValue = 2;
Chris@188:     desc.defaultValue = float(m_defaultDParams.metric);
Chris@156:     desc.isQuantized = true;
Chris@156:     desc.quantizeStep = 1;
Chris@156:     desc.valueNames.clear();
Chris@157:     desc.valueNames.push_back("Manhattan");
Chris@156:     desc.valueNames.push_back("Euclidean");
Chris@156:     desc.valueNames.push_back("Cosine");
Chris@156:     list.push_back(desc);
Chris@156:     desc.valueNames.clear();
Chris@156: 
Chris@162:     desc.identifier = "distnorm";
Chris@162:     desc.name = "Distance normalisation";
Chris@162:     desc.description = "Type of normalisation to use for distance metric";
Chris@162:     desc.minValue = 0;
Chris@162:     desc.maxValue = 2;
Chris@188:     desc.defaultValue = float(m_defaultDParams.norm);
Chris@162:     desc.isQuantized = true;
Chris@162:     desc.quantizeStep = 1;
Chris@162:     desc.valueNames.clear();
Chris@162:     desc.valueNames.push_back("None");
Chris@162:     desc.valueNames.push_back("Sum of frames");
Chris@162:     desc.valueNames.push_back("Log sum of frames");
Chris@162:     list.push_back(desc);
Chris@162:     desc.valueNames.clear();
Chris@162: 
Chris@198: #ifdef USE_COMPACT_TYPES
Chris@198:     desc.identifier = "scale";
Chris@198:     desc.name = "Distance scale";
Chris@198:     desc.description = "Scale factor to use when mapping distance metric into byte range for storage";
Chris@198:     desc.minValue = 1;
Chris@198:     desc.maxValue = 1000;
Chris@198:     desc.defaultValue = float(m_defaultDParams.scale);
Chris@198:     desc.isQuantized = false;
Chris@198:     list.push_back(desc);
Chris@198: #endif
Chris@198:     
Chris@162:     desc.identifier = "silencethreshold";
Chris@162:     desc.name = "Silence threshold";
Chris@162:     desc.description = "Total frame energy threshold below which a feature will be regarded as silent";
Chris@162:     desc.minValue = 0;
Chris@165:     desc.maxValue = 0.1f;
Chris@188:     desc.defaultValue = float(m_defaultFcParams.silenceThreshold);
Chris@162:     desc.isQuantized = false;
Chris@162:     list.push_back(desc);
Chris@162: 
Chris@151:     desc.identifier = "noise";
Chris@162:     desc.name = "Add noise";
Chris@241:     desc.description = "Whether to mix in a small constant white noise term when calculating feature distance. This can improve alignment against sources containing cleanly synthesised audio";
Chris@151:     desc.minValue = 0;
Chris@151:     desc.maxValue = 1;
Chris@188:     desc.defaultValue = float(m_defaultDParams.noise);
Chris@151:     desc.isQuantized = true;
Chris@151:     desc.quantizeStep = 1;
Chris@151:     list.push_back(desc);
Chris@136:     
Chris@25:     desc.identifier = "gradientlimit";
Chris@162:     desc.name = "Gradient limit";
Chris@18:     desc.description = "Limit of number of frames that will be accepted from one source without a frame from the other source being accepted";
Chris@18:     desc.minValue = 1;
Chris@18:     desc.maxValue = 10;
Chris@188:     desc.defaultValue = float(m_defaultParams.maxRunCount);
Chris@18:     desc.isQuantized = true;
Chris@18:     desc.quantizeStep = 1;
Chris@18:     list.push_back(desc);
Chris@18: 
Chris@25:     desc.identifier = "zonewidth";
Chris@162:     desc.name = "Search zone width";
Chris@25:     desc.description = "Width of the search zone (error margin) either side of the ongoing match position, in seconds";
Chris@25:     desc.minValue = 1;
Chris@25:     desc.maxValue = 60;
Chris@188:     desc.defaultValue = float(m_defaultParams.blockTime);
Chris@25:     desc.isQuantized = true;
Chris@25:     desc.quantizeStep = 1;
Chris@25:     desc.unit = "s";
Chris@25:     list.push_back(desc);
Chris@25: 
Chris@83:     desc.identifier = "diagonalweight";
Chris@162:     desc.name = "Diagonal weight";
Chris@83:     desc.description = "Weight applied to cost of diagonal step relative to horizontal or vertical step. The default of 2.0 is good for gross tracking of quite different performances; closer to 1.0 produces a smoother path for performances more similar in tempo";
Chris@83:     desc.minValue = 1.0;
Chris@86:     desc.maxValue = 2.0;
Chris@188:     desc.defaultValue = float(m_defaultParams.diagonalWeight);
Chris@83:     desc.isQuantized = false;
Chris@83:     desc.unit = "";
Chris@83:     list.push_back(desc);
Chris@83:     
Chris@32:     desc.identifier = "smooth";
Chris@162:     desc.name = "Use path smoothing";
Chris@138:     desc.description = "Smooth the path by replacing steps with diagonals. (This was enabled by default in earlier versions of the MATCH plugin, but the default now is to produce an un-smoothed path.)";
Chris@32:     desc.minValue = 0;
Chris@32:     desc.maxValue = 1;
Chris@138:     desc.defaultValue = 0;
Chris@32:     desc.isQuantized = true;
Chris@32:     desc.quantizeStep = 1;
Chris@32:     desc.unit = "";
Chris@32:     list.push_back(desc);
Chris@32: 
Chris@162:     desc.identifier = "serialise";
Chris@162:     desc.name = "Serialise plugin invocations";
Chris@162:     desc.description = "Reduce potential memory load at the expense of multiprocessor performance by serialising multi-threaded plugin runs";
Chris@162:     desc.minValue = 0;
Chris@162:     desc.maxValue = 1;
Chris@162:     desc.defaultValue = 0;
Chris@162:     desc.isQuantized = true;
Chris@162:     desc.quantizeStep = 1;
Chris@161:     list.push_back(desc);
Chris@236: 
cannam@0:     return list;
cannam@0: }
cannam@0: 
cannam@0: float
cannam@0: MatchVampPlugin::getParameter(std::string name) const
cannam@0: {
cannam@0:     if (name == "serialise") {
cannam@0:         return m_serialise ? 1.0 : 0.0; 
Chris@18:     } else if (name == "framenorm") {
Chris@188:         return float(m_fcParams.norm);
Chris@18:     } else if (name == "distnorm") {
Chris@188:         return float(m_dParams.norm);
Chris@18:     } else if (name == "usespecdiff") {
Chris@188:         return float(m_fcParams.order);
Chris@18:     } else if (name == "usechroma") {
Chris@38:         return m_feParams.useChromaFrequencyMap ? 1.0 : 0.0;
Chris@25:     } else if (name == "gradientlimit") {
Chris@188:         return float(m_params.maxRunCount);
Chris@83:     } else if (name == "diagonalweight") {
Chris@188:         return float(m_params.diagonalWeight);
Chris@25:     } else if (name == "zonewidth") {
Chris@188:         return float(m_params.blockTime);
Chris@32:     } else if (name == "smooth") {
Chris@32:         return m_smooth ? 1.0 : 0.0;
Chris@136:     } else if (name == "silencethreshold") {
Chris@188:         return float(m_fcParams.silenceThreshold);
Chris@156:     } else if (name == "metric") {
Chris@188:         return float(m_dParams.metric);
Chris@151:     } else if (name == "noise") {
Chris@151:         return m_dParams.noise;
Chris@198:     } else if (name == "scale") {
Chris@198:         return float(m_dParams.scale);
Chris@161:     } else if (name == "freq1") {
Chris@188:         return float(m_feParams.referenceFrequency);
Chris@161:     } else if (name == "freq2") {
Chris@188:         return float(m_secondReferenceFrequency);
Chris@177:     } else if (name == "minfreq") {
Chris@188:         return float(m_feParams.minFrequency);
Chris@177:     } else if (name == "maxfreq") {
Chris@188:         return float(m_feParams.maxFrequency);
cannam@0:     }
Chris@18:     
cannam@0:     return 0.0;
cannam@0: }
cannam@0: 
cannam@0: void
cannam@0: MatchVampPlugin::setParameter(std::string name, float value)
cannam@0: {
cannam@0:     if (name == "serialise") {
cannam@0:         m_serialise = (value > 0.5);
Chris@18:     } else if (name == "framenorm") {
Chris@188:         m_fcParams.norm = FeatureConditioner::Normalisation(int(value + 0.1));
Chris@18:     } else if (name == "distnorm") {
Chris@188:         m_dParams.norm = DistanceMetric::DistanceNormalisation(int(value + 0.1));
Chris@18:     } else if (name == "usespecdiff") {
Chris@188:         m_fcParams.order = FeatureConditioner::OutputOrder(int(value + 0.1));
Chris@18:     } else if (name == "usechroma") {
Chris@38:         m_feParams.useChromaFrequencyMap = (value > 0.5);
Chris@25:     } else if (name == "gradientlimit") {
Chris@18:         m_params.maxRunCount = int(value + 0.1);
Chris@83:     } else if (name == "diagonalweight") {
Chris@83:         m_params.diagonalWeight = value;
Chris@25:     } else if (name == "zonewidth") {
Chris@25:         m_params.blockTime = value;
Chris@32:     } else if (name == "smooth") {
Chris@32:         m_smooth = (value > 0.5);
Chris@136:     } else if (name == "silencethreshold") {
Chris@136:         m_fcParams.silenceThreshold = value;
Chris@156:     } else if (name == "metric") {
Chris@188:         m_dParams.metric = DistanceMetric::Metric(int(value + 0.1));
Chris@151:     } else if (name == "noise") {
Chris@188:         m_dParams.noise = DistanceMetric::NoiseAddition(int(value + 0.1));
Chris@198:     } else if (name == "scale") {
Chris@198:         m_dParams.scale = value;
Chris@161:     } else if (name == "freq1") {
Chris@161:         m_feParams.referenceFrequency = value;
Chris@161:     } else if (name == "freq2") {
Chris@161:         m_secondReferenceFrequency = value;
Chris@177:     } else if (name == "minfreq") {
Chris@177:         m_feParams.minFrequency = value;
Chris@177:     } else if (name == "maxfreq") {
Chris@177:         m_feParams.maxFrequency = value;
cannam@0:     }
cannam@0: }
cannam@0: 
cannam@0: size_t
cannam@0: MatchVampPlugin::getPreferredStepSize() const
cannam@0: {
Chris@52:     return int(m_inputSampleRate * defaultStepTime + 0.001);
cannam@0: }
cannam@0: 
cannam@0: size_t
cannam@0: MatchVampPlugin::getPreferredBlockSize() const
cannam@0: {
Chris@216:     return m_defaultFeParams.fftSize;
cannam@0: }
cannam@0: 
cannam@0: void
Chris@17: MatchVampPlugin::createMatchers()
cannam@0: {
Chris@17:     m_params.hopTime = m_stepTime;
Chris@38:     m_feParams.fftSize = m_blockSize;
Chris@107: 
Chris@161:     m_pipeline = new MatchPipeline(m_feParams, m_fcParams, m_dParams, m_params,
Chris@161:                                    m_secondReferenceFrequency);
cannam@0: }
cannam@0: 
cannam@0: bool
cannam@0: MatchVampPlugin::initialise(size_t channels, size_t stepSize, size_t blockSize)
cannam@0: {
Chris@10:     if (m_inputSampleRate < sampleRateMin) {
Chris@10:         std::cerr << "MatchVampPlugin::MatchVampPlugin: input sample rate "
Chris@10:                   << m_inputSampleRate << " < min supported rate "
Chris@10:                   << sampleRateMin << std::endl;
Chris@10:         return false;
Chris@10:     }
cannam@0:     if (channels < getMinChannelCount() ||
cannam@0: 	channels > getMaxChannelCount()) return false;
cannam@1:     if (stepSize > blockSize/2 ||
cannam@0:         blockSize != getPreferredBlockSize()) return false;
Chris@15: 
Chris@188:     m_stepSize = int(stepSize);
Chris@15:     m_stepTime = float(stepSize) / m_inputSampleRate;
Chris@188:     m_blockSize = int(blockSize);
Chris@15: 
Chris@15:     createMatchers();
cannam@0:     m_begin = true;
cannam@0:     m_locked = false;
Chris@15: 
cannam@0:     return true;
cannam@0: }
cannam@0: 
cannam@0: void
cannam@0: MatchVampPlugin::reset()
cannam@0: {
Chris@107:     delete m_pipeline;
Chris@107:     m_pipeline = 0;
Chris@74:     m_frameNo = 0;
cannam@6:     createMatchers();
cannam@6:     m_begin = true;
cannam@6:     m_locked = false;
cannam@0: }
cannam@0: 
cannam@0: MatchVampPlugin::OutputList
cannam@0: MatchVampPlugin::getOutputDescriptors() const
cannam@0: {
cannam@0:     OutputList list;
cannam@0: 
Chris@52:     float outRate = 1.0f / m_stepTime;
cannam@0: 
cannam@0:     OutputDescriptor desc;
cannam@0:     desc.identifier = "path";
cannam@0:     desc.name = "Path";
cannam@0:     desc.description = "Alignment path";
cannam@0:     desc.unit = "";
cannam@0:     desc.hasFixedBinCount = true;
cannam@0:     desc.binCount = 1;
cannam@0:     desc.hasKnownExtents = false;
cannam@0:     desc.isQuantized = true;
cannam@0:     desc.quantizeStep = 1;
cannam@0:     desc.sampleType = OutputDescriptor::VariableSampleRate;
cannam@0:     desc.sampleRate = outRate;
Chris@180:     m_pathOutNo = int(list.size());
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc.identifier = "a_b";
cannam@0:     desc.name = "A-B Timeline";
cannam@0:     desc.description = "Timing in performance B corresponding to moments in performance A";
cannam@0:     desc.unit = "sec";
cannam@0:     desc.hasFixedBinCount = true;
cannam@0:     desc.binCount = 1;
cannam@0:     desc.hasKnownExtents = false;
cannam@0:     desc.isQuantized = false;
cannam@0:     desc.sampleType = OutputDescriptor::VariableSampleRate;
cannam@0:     desc.sampleRate = outRate;
Chris@180:     m_abOutNo = int(list.size());
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc.identifier = "b_a";
cannam@0:     desc.name = "B-A Timeline";
cannam@0:     desc.description = "Timing in performance A corresponding to moments in performance B";
cannam@0:     desc.unit = "sec";
cannam@0:     desc.hasFixedBinCount = true;
cannam@0:     desc.binCount = 1;
cannam@0:     desc.hasKnownExtents = false;
cannam@0:     desc.isQuantized = false;
cannam@0:     desc.sampleType = OutputDescriptor::VariableSampleRate;
cannam@0:     desc.sampleRate = outRate;
Chris@180:     m_baOutNo = int(list.size());
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc.identifier = "a_b_divergence";
cannam@0:     desc.name = "A-B Divergence";
cannam@0:     desc.description = "Difference between timings in performances A and B";
cannam@0:     desc.unit = "sec";
cannam@0:     desc.hasFixedBinCount = true;
cannam@0:     desc.binCount = 1;
cannam@0:     desc.hasKnownExtents = false;
cannam@0:     desc.isQuantized = false;
cannam@0:     desc.sampleType = OutputDescriptor::VariableSampleRate;
cannam@0:     desc.sampleRate = outRate;
Chris@180:     m_abDivOutNo = int(list.size());
cannam@0:     list.push_back(desc);
cannam@0: 
cannam@0:     desc.identifier = "a_b_temporatio";
cannam@0:     desc.name = "A-B Tempo Ratio";
cannam@0:     desc.description = "Ratio of tempi between performances A and B";
cannam@0:     desc.unit = "";
cannam@0:     desc.hasFixedBinCount = true;
cannam@0:     desc.binCount = 1;
cannam@0:     desc.hasKnownExtents = false;
cannam@0:     desc.isQuantized = false;
cannam@0:     desc.sampleType = OutputDescriptor::VariableSampleRate;
cannam@0:     desc.sampleRate = outRate;
Chris@180:     m_abRatioOutNo = int(list.size());
cannam@0:     list.push_back(desc);
cannam@0: 
Chris@38:     int featureSize = FeatureExtractor(m_feParams).getFeatureSize();
Chris@38:     
Chris@15:     desc.identifier = "a_features";
Chris@140:     desc.name = "Raw A Features";
Chris@15:     desc.description = "Spectral features extracted from performance A";
Chris@15:     desc.unit = "";
Chris@15:     desc.hasFixedBinCount = true;
Chris@38:     desc.binCount = featureSize;
Chris@15:     desc.hasKnownExtents = false;
Chris@15:     desc.isQuantized = false;
Chris@16:     desc.sampleType = OutputDescriptor::FixedSampleRate;
Chris@15:     desc.sampleRate = outRate;
Chris@180:     m_aFeaturesOutNo = int(list.size());
Chris@16:     list.push_back(desc);
Chris@16: 
Chris@16:     desc.identifier = "b_features";
Chris@140:     desc.name = "Raw B Features";
Chris@16:     desc.description = "Spectral features extracted from performance B";
Chris@16:     desc.unit = "";
Chris@16:     desc.hasFixedBinCount = true;
Chris@38:     desc.binCount = featureSize;
Chris@16:     desc.hasKnownExtents = false;
Chris@16:     desc.isQuantized = false;
Chris@16:     desc.sampleType = OutputDescriptor::FixedSampleRate;
Chris@16:     desc.sampleRate = outRate;
Chris@180:     m_bFeaturesOutNo = int(list.size());
Chris@15:     list.push_back(desc);
Chris@15: 
Chris@140:     desc.identifier = "a_cfeatures";
Chris@140:     desc.name = "Conditioned A Features";
Chris@140:     desc.description = "Spectral features extracted from performance A, after normalisation and conditioning";
Chris@140:     desc.unit = "";
Chris@140:     desc.hasFixedBinCount = true;
Chris@140:     desc.binCount = featureSize;
Chris@140:     desc.hasKnownExtents = false;
Chris@140:     desc.isQuantized = false;
Chris@140:     desc.sampleType = OutputDescriptor::FixedSampleRate;
Chris@140:     desc.sampleRate = outRate;
Chris@180:     m_caFeaturesOutNo = int(list.size());
Chris@140:     list.push_back(desc);
Chris@140: 
Chris@140:     desc.identifier = "b_cfeatures";
Chris@140:     desc.name = "Conditioned B Features";
Chris@140:     desc.description = "Spectral features extracted from performance B, after norrmalisation and conditioning";
Chris@140:     desc.unit = "";
Chris@140:     desc.hasFixedBinCount = true;
Chris@140:     desc.binCount = featureSize;
Chris@140:     desc.hasKnownExtents = false;
Chris@140:     desc.isQuantized = false;
Chris@140:     desc.sampleType = OutputDescriptor::FixedSampleRate;
Chris@140:     desc.sampleRate = outRate;
Chris@180:     m_cbFeaturesOutNo = int(list.size());
Chris@140:     list.push_back(desc);
Chris@140: 
Chris@163:     desc.identifier = "overall_cost";
Chris@163:     desc.name = "Overall Cost";
Chris@163:     desc.description = "Normalised overall path cost for the cheapest path";
Chris@163:     desc.unit = "";
Chris@163:     desc.hasFixedBinCount = true;
Chris@163:     desc.binCount = 1;
Chris@163:     desc.hasKnownExtents = false;
Chris@163:     desc.isQuantized = false;
Chris@163:     desc.sampleType = OutputDescriptor::FixedSampleRate;
Chris@163:     desc.sampleRate = 1;
Chris@180:     m_overallCostOutNo = int(list.size());
Chris@163:     list.push_back(desc);
Chris@163:     
cannam@0:     return list;
cannam@0: }
cannam@0: 
cannam@0: MatchVampPlugin::FeatureSet
cannam@0: MatchVampPlugin::process(const float *const *inputBuffers,
cannam@0:                          Vamp::RealTime timestamp)
cannam@0: {
cannam@0:     if (m_begin) {
cannam@0:         if (!m_locked && m_serialise) {
cannam@0:             m_locked = true;
cannam@0: #ifdef _WIN32
cannam@0:             WaitForSingleObject(m_serialisingMutex, INFINITE);
cannam@0: #else
cannam@0:             pthread_mutex_lock(&m_serialisingMutex);
cannam@0: #endif
cannam@0:         }
Chris@10:         m_startTime = timestamp;
cannam@0:         m_begin = false;
cannam@0:     }
cannam@0:     
cannam@0: //    std::cerr << timestamp.toString();
cannam@0: 
Chris@107:     m_pipeline->feedFrequencyDomainAudio(inputBuffers[0], inputBuffers[1]);
Chris@74: 
Chris@140:     FeatureSet returnFeatures;
Chris@140: 
Chris@185:     feature_t f1, f2;
Chris@140:     m_pipeline->extractFeatures(f1, f2);
Chris@16: 
Chris@185:     feature_t cf1, cf2;
Chris@140:     m_pipeline->extractConditionedFeatures(cf1, cf2);
Chris@16: 
Chris@16:     Feature f;
Chris@16:     f.hasTimestamp = false;
Chris@16: 
Chris@74:     f.values.clear();
Chris@188:     for (auto v: f1) f.values.push_back(float(v));
Chris@74:     returnFeatures[m_aFeaturesOutNo].push_back(f);
Chris@16: 
Chris@74:     f.values.clear();
Chris@188:     for (auto v: f2) f.values.push_back(float(v));
Chris@74:     returnFeatures[m_bFeaturesOutNo].push_back(f);
cannam@0: 
Chris@140:     f.values.clear();
Chris@188:     for (auto v: cf1) f.values.push_back(float(v));
Chris@140:     returnFeatures[m_caFeaturesOutNo].push_back(f);
Chris@140: 
Chris@140:     f.values.clear();
Chris@188:     for (auto v: cf2) f.values.push_back(float(v));
Chris@140:     returnFeatures[m_cbFeaturesOutNo].push_back(f);
Chris@140: 
cannam@0: //    std::cerr << ".";
cannam@0: //    std::cerr << std::endl;
cannam@0: 
Chris@74:     ++m_frameNo;
Chris@74:     
Chris@16:     return returnFeatures;
cannam@0: }
cannam@0: 
cannam@0: MatchVampPlugin::FeatureSet
cannam@0: MatchVampPlugin::getRemainingFeatures()
cannam@0: {
Chris@107:     m_pipeline->finish();
Chris@74: 
Chris@63:     FeatureSet returnFeatures;
Chris@63:     
cannam@0:     std::vector<int> pathx;
cannam@0:     std::vector<int> pathy;
Chris@155:     int len = m_pipeline->retrievePath(m_smooth, pathx, pathy);
cannam@0: 
Chris@173:     double cost = m_pipeline->getOverallCost();
Chris@163:     Feature costFeature;
Chris@163:     costFeature.hasTimestamp = false;
Chris@188:     costFeature.values.push_back(float(cost));
Chris@163:     returnFeatures[m_overallCostOutNo].push_back(costFeature);
Chris@163:     
cannam@0:     int prevx = 0;
cannam@0:     int prevy = 0;
cannam@0: 
Chris@30:     for (int i = 0; i < len; ++i) {
cannam@0: 
cannam@0:         int x = pathx[i];
cannam@0:         int y = pathy[i];
cannam@0: 
cannam@0:         Vamp::RealTime xt = Vamp::RealTime::frame2RealTime
Chris@180:             (x * m_stepSize, int(m_inputSampleRate + 0.5));
cannam@0:         Vamp::RealTime yt = Vamp::RealTime::frame2RealTime
Chris@180:             (y * m_stepSize, int(m_inputSampleRate + 0.5));
cannam@0: 
cannam@0:         Feature feature;
cannam@0:         feature.hasTimestamp = true;
Chris@10:         feature.timestamp = m_startTime + xt;
cannam@0:         feature.values.clear();
Chris@52:         feature.values.push_back(float(yt.sec + double(yt.nsec)/1.0e9));
Chris@16:         returnFeatures[m_pathOutNo].push_back(feature);
cannam@0:         
cannam@0:         if (x != prevx) {
cannam@0: 
cannam@0:             feature.hasTimestamp = true;
Chris@10:             feature.timestamp = m_startTime + xt;
cannam@0:             feature.values.clear();
Chris@52:             feature.values.push_back(float(yt.sec + yt.msec()/1000.0));
Chris@16:             returnFeatures[m_abOutNo].push_back(feature);
cannam@0: 
cannam@0:             Vamp::RealTime diff = yt - xt;
cannam@0:             feature.values.clear();
Chris@52:             feature.values.push_back(float(diff.sec + diff.msec()/1000.0));
Chris@16:             returnFeatures[m_abDivOutNo].push_back(feature);
cannam@0: 
cannam@0:             if (i > 0) {
cannam@0:                 int lookback = 100; //!!! arbitrary
cannam@0:                 if (lookback > i) lookback = i;
cannam@0:                 int xdiff = x - pathx[i-lookback];
cannam@0:                 int ydiff = y - pathy[i-lookback];
cannam@0:                 if (xdiff != 0 && ydiff != 0) {
cannam@0:                     float ratio = float(ydiff)/float(xdiff);
cannam@0:                     if (ratio < 8 && ratio > (1.0/8)) { //!!! just for now, since we aren't dealing properly with silence yet
cannam@0:                         feature.values.clear();
cannam@0:                         feature.values.push_back(ratio);
Chris@16:                         returnFeatures[m_abRatioOutNo].push_back(feature);
cannam@0:                     }
cannam@0:                 }
cannam@0:             }
cannam@0:         }
cannam@0: 
cannam@0:         if (y != prevy) {
cannam@0:             feature.hasTimestamp = true;
Chris@10:             feature.timestamp = m_startTime + yt;
cannam@0:             feature.values.clear();
Chris@52:             feature.values.push_back(float(xt.sec + xt.msec()/1000.0));
Chris@16:             returnFeatures[m_baOutNo].push_back(feature);
cannam@0:         }
cannam@0: 
cannam@0:         prevx = x;
cannam@0:         prevy = y;
cannam@0:     }
cannam@0: 
Chris@107:     delete m_pipeline;
Chris@107:     m_pipeline = 0;
cannam@0: 
cannam@0:     if (m_locked) {
cannam@0: #ifdef _WIN32
cannam@0:         ReleaseMutex(m_serialisingMutex);
cannam@0: #else
cannam@0:         pthread_mutex_unlock(&m_serialisingMutex);
cannam@0: #endif
cannam@0:         m_locked = false;
cannam@0:     }
cannam@0: 
cannam@0:     return returnFeatures;
cannam@0:     
cannam@0: 
cannam@0: /*
Chris@30:     for (int i = 0; i < len; ++i) {
cannam@0:         std::cerr << i << ": [" << pathx[i] << "," << pathy[i] << "]" << std::endl;
cannam@0:     }
cannam@0: 
cannam@0:     std::cerr << std::endl;
cannam@0:     std::cerr << "File: A" << std::endl;
cannam@0:     std::cerr << "Marks: -1" << std::endl;
cannam@0:     std::cerr << "FixedPoints: true 0" << std::endl;
cannam@0:     std::cerr << "0" << std::endl;
cannam@0:     std::cerr << "0" << std::endl;
cannam@0:     std::cerr << "0" << std::endl;
cannam@0:     std::cerr << "0" << std::endl;
cannam@0:     std::cerr << "File: B" << std::endl;
cannam@0:     std::cerr << "Marks: 0" << std::endl;
cannam@0:     std::cerr << "FixedPoints: true 0" << std::endl;
cannam@0:     std::cerr << "0.02" << std::endl;
cannam@0:     std::cerr << "0.02" << std::endl;
cannam@0: 
Chris@30:     std::cerr << len << std::endl;
Chris@30:     for (int i = 0; i < len; ++i) {
cannam@0:         std::cerr << pathx[i] << std::endl;
cannam@0:     }
cannam@0: 
Chris@30:     std::cerr << len << std::endl;
Chris@30:     for (int i = 0; i < len; ++i) {
cannam@0:         std::cerr << pathy[i] << std::endl;
cannam@0:     }
cannam@0: */
cannam@0: }