Mercurial > hg > match-vamp
changeset 173:eeed3498fe96 refactors
Merge from branch structure
| author | Chris Cannam |
|---|---|
| date | Fri, 06 Feb 2015 18:41:35 +0000 |
| parents | d23dad16d6f9 (diff) 30d59e1e4232 (current diff) |
| children | ef3c4b451c57 |
| files | src/Finder.cpp src/Finder.h src/MatchPipeline.cpp src/MatchPipeline.h src/MatchVampPlugin.cpp |
| diffstat | 9 files changed, 421 insertions(+), 85 deletions(-) [+] |
line wrap: on
line diff
--- a/match-vamp-plugin.n3 Fri Feb 06 18:09:18 2015 +0000 +++ b/match-vamp-plugin.n3 Fri Feb 06 18:41:35 2015 +0000 @@ -6,24 +6,26 @@ @prefix dc: <http://purl.org/dc/elements/1.1/> . @prefix af: <http://purl.org/ontology/af/> . @prefix foaf: <http://xmlns.com/foaf/0.1/> . +@prefix doap: <http://usefulinc.com/ns/doap#> . @prefix cc: <http://web.resource.org/cc/> . @prefix : <#> . -<> a vamp:PluginDescription ; - foaf:maker <http://www.vamp-plugins.org/doap.rdf#template-generator> ; - foaf:primaryTopic <http://vamp-plugins.org/rdf/plugins/match-vamp-plugin> . +<> a vamp:PluginDescription ; + foaf:maker <http://vamp-plugins.org/rdf/template-generator> ; + foaf:primaryTopic <http://vamp-plugins.org/rdf/plugins/match-vamp-plugin> . -:maker +:library_maker foaf:name "Simon Dixon and Chris Cannam" ; foaf:logo <http://vamp-plugins.org/rdf/plugins/makers/qm.png> ; - foaf:page <http://c4dm.eecs.qmul.ac.uk/> . + foaf:page <http://c4dm.eecs.qmul.ac.uk/> ; + . plugbase:library a vamp:PluginLibrary ; vamp:identifier "match-vamp-plugin" ; dc:title "MATCH Vamp Plugin" ; dc:description """Vamp implementation of the MATCH audio alignment algorithm from Simon Dixon. Sonic Visualiser can use this for automatic time alignment among multiple audio files.""" ; vamp:available_plugin plugbase:match ; - foaf:maker :maker ; + foaf:maker :library_maker ; foaf:page <http://code.soundsoftware.ac.uk/projects/match-vamp> ; . @@ -31,15 +33,27 @@ dc:title "Match Performance Aligner" ; vamp:name "Match Performance Aligner" ; dc:description """Calculate alignment between two performances in separate channel inputs""" ; - foaf:maker :maker ; + foaf:maker :library_maker ; dc:rights """GPL""" ; # cc:license <Place plugin license URI here and uncomment> ; vamp:identifier "match" ; vamp:vamp_API_version vamp:api_version_2 ; - owl:versionInfo "1" ; + owl:versionInfo "3" ; vamp:input_domain vamp:FrequencyDomain ; - + vamp:parameter plugbase:match_param_freq1 ; + vamp:parameter plugbase:match_param_freq2 ; + vamp:parameter plugbase:match_param_usechroma ; + vamp:parameter plugbase:match_param_usespecdiff ; + vamp:parameter plugbase:match_param_framenorm ; + vamp:parameter plugbase:match_param_metric ; + vamp:parameter plugbase:match_param_distnorm ; + vamp:parameter plugbase:match_param_silencethreshold ; + vamp:parameter plugbase:match_param_noise ; + vamp:parameter plugbase:match_param_gradientlimit ; + vamp:parameter plugbase:match_param_zonewidth ; + vamp:parameter plugbase:match_param_diagonalweight ; + vamp:parameter plugbase:match_param_smooth ; vamp:parameter plugbase:match_param_serialise ; vamp:output plugbase:match_output_path ; @@ -47,10 +61,155 @@ vamp:output plugbase:match_output_b_a ; vamp:output plugbase:match_output_a_b_divergence ; vamp:output plugbase:match_output_a_b_temporatio ; + vamp:output plugbase:match_output_a_features ; + vamp:output plugbase:match_output_b_features ; + vamp:output plugbase:match_output_a_cfeatures ; + vamp:output plugbase:match_output_b_cfeatures ; + vamp:output plugbase:match_output_overall_cost ; + . + +plugbase:match_param_freq1 a vamp:Parameter ; + vamp:identifier "freq1" ; + dc:title "Tuning frequency of first input" ; + dc:format "Hz" ; + vamp:min_value 220 ; + vamp:max_value 880 ; + vamp:unit "Hz" ; + vamp:default_value 440 ; + vamp:value_names (); + . +plugbase:match_param_freq2 a vamp:Parameter ; + vamp:identifier "freq2" ; + dc:title "Tuning frequency of second input" ; + dc:format "Hz" ; + vamp:min_value 220 ; + vamp:max_value 880 ; + vamp:unit "Hz" ; + vamp:default_value 440 ; + vamp:value_names (); + . +plugbase:match_param_usechroma a vamp:QuantizedParameter ; + vamp:identifier "usechroma" ; + dc:title "Feature type" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 1 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 0 ; + vamp:value_names ( "Spectral" "Chroma"); + . +plugbase:match_param_usespecdiff a vamp:QuantizedParameter ; + vamp:identifier "usespecdiff" ; + dc:title "Use feature difference" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 1 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 1 ; + vamp:value_names (); + . +plugbase:match_param_framenorm a vamp:QuantizedParameter ; + vamp:identifier "framenorm" ; + dc:title "Frame normalisation" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 2 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 1 ; + vamp:value_names ( "None" "Sum to 1" "Long-term average"); + . +plugbase:match_param_metric a vamp:QuantizedParameter ; + vamp:identifier "metric" ; + dc:title "Distance metric" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 2 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 0 ; + vamp:value_names ( "Manhattan" "Euclidean" "Cosine"); + . +plugbase:match_param_distnorm a vamp:QuantizedParameter ; + vamp:identifier "distnorm" ; + dc:title "Distance normalisation" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 2 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 2 ; + vamp:value_names ( "None" "Sum of frames" "Log sum of frames"); + . +plugbase:match_param_silencethreshold a vamp:Parameter ; + vamp:identifier "silencethreshold" ; + dc:title "Silence threshold" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 0.1 ; + vamp:unit "" ; + vamp:default_value 0.01 ; + vamp:value_names (); + . +plugbase:match_param_noise a vamp:QuantizedParameter ; + vamp:identifier "noise" ; + dc:title "Add noise" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 1 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 1 ; + vamp:value_names (); + . +plugbase:match_param_gradientlimit a vamp:QuantizedParameter ; + vamp:identifier "gradientlimit" ; + dc:title "Gradient limit" ; + dc:format "" ; + vamp:min_value 1 ; + vamp:max_value 10 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 3 ; + vamp:value_names (); + . +plugbase:match_param_zonewidth a vamp:QuantizedParameter ; + vamp:identifier "zonewidth" ; + dc:title "Search zone width" ; + dc:format "s" ; + vamp:min_value 1 ; + vamp:max_value 60 ; + vamp:unit "s" ; + vamp:quantize_step 1 ; + vamp:default_value 10 ; + vamp:value_names (); + . +plugbase:match_param_diagonalweight a vamp:Parameter ; + vamp:identifier "diagonalweight" ; + dc:title "Diagonal weight" ; + dc:format "" ; + vamp:min_value 1 ; + vamp:max_value 2 ; + vamp:unit "" ; + vamp:default_value 2 ; + vamp:value_names (); + . +plugbase:match_param_smooth a vamp:QuantizedParameter ; + vamp:identifier "smooth" ; + dc:title "Use path smoothing" ; + dc:format "" ; + vamp:min_value 0 ; + vamp:max_value 1 ; + vamp:unit "" ; + vamp:quantize_step 1 ; + vamp:default_value 0 ; + vamp:value_names (); . plugbase:match_param_serialise a vamp:QuantizedParameter ; vamp:identifier "serialise" ; - dc:title "Serialise Plugin Invocations" ; + dc:title "Serialise plugin invocations" ; dc:format "" ; vamp:min_value 0 ; vamp:max_value 1 ; @@ -62,7 +221,7 @@ plugbase:match_output_path a vamp:SparseOutput ; vamp:identifier "path" ; dc:title "Path" ; - dc:description "Alignment path" ; + dc:description """Alignment path""" ; vamp:fixed_bin_count "true" ; vamp:unit "" ; a vamp:QuantizedOutput ; @@ -77,7 +236,7 @@ plugbase:match_output_a_b a vamp:SparseOutput ; vamp:identifier "a_b" ; dc:title "A-B Timeline" ; - dc:description "Timing in performance B corresponding to moments in performance A" ; + dc:description """Timing in performance B corresponding to moments in performance A""" ; vamp:fixed_bin_count "true" ; vamp:unit "sec" ; vamp:bin_count 1 ; @@ -90,7 +249,7 @@ plugbase:match_output_b_a a vamp:SparseOutput ; vamp:identifier "b_a" ; dc:title "B-A Timeline" ; - dc:description "Timing in performance A corresponding to moments in performance B" ; + dc:description """Timing in performance A corresponding to moments in performance B""" ; vamp:fixed_bin_count "true" ; vamp:unit "sec" ; vamp:bin_count 1 ; @@ -103,7 +262,7 @@ plugbase:match_output_a_b_divergence a vamp:SparseOutput ; vamp:identifier "a_b_divergence" ; dc:title "A-B Divergence" ; - dc:description "Difference between timings in performances A and B" ; + dc:description """Difference between timings in performances A and B""" ; vamp:fixed_bin_count "true" ; vamp:unit "sec" ; vamp:bin_count 1 ; @@ -116,7 +275,7 @@ plugbase:match_output_a_b_temporatio a vamp:SparseOutput ; vamp:identifier "a_b_temporatio" ; dc:title "A-B Tempo Ratio" ; - dc:description "Ratio of tempi between performances A and B" ; + dc:description """Ratio of tempi between performances A and B""" ; vamp:fixed_bin_count "true" ; vamp:unit "" ; vamp:bin_count 1 ; @@ -126,4 +285,59 @@ # vamp:computes_feature <Place feature attribute URI here and uncomment> ; # vamp:computes_signal_type <Place signal type URI here and uncomment> ; . +plugbase:match_output_a_features a vamp:DenseOutput ; + vamp:identifier "a_features" ; + dc:title "Raw A Features" ; + dc:description """Spectral features extracted from performance A""" ; + vamp:fixed_bin_count "true" ; + vamp:unit "" ; + vamp:bin_count 84 ; +# vamp:computes_event_type <Place event type URI here and uncomment> ; +# vamp:computes_feature <Place feature attribute URI here and uncomment> ; +# vamp:computes_signal_type <Place signal type URI here and uncomment> ; + . +plugbase:match_output_b_features a vamp:DenseOutput ; + vamp:identifier "b_features" ; + dc:title "Raw B Features" ; + dc:description """Spectral features extracted from performance B""" ; + vamp:fixed_bin_count "true" ; + vamp:unit "" ; + vamp:bin_count 84 ; +# vamp:computes_event_type <Place event type URI here and uncomment> ; +# vamp:computes_feature <Place feature attribute URI here and uncomment> ; +# vamp:computes_signal_type <Place signal type URI here and uncomment> ; + . +plugbase:match_output_a_cfeatures a vamp:DenseOutput ; + vamp:identifier "a_cfeatures" ; + dc:title "Conditioned A Features" ; + dc:description """Spectral features extracted from performance A, after normalisation and conditioning""" ; + vamp:fixed_bin_count "true" ; + vamp:unit "" ; + vamp:bin_count 84 ; +# vamp:computes_event_type <Place event type URI here and uncomment> ; +# vamp:computes_feature <Place feature attribute URI here and uncomment> ; +# vamp:computes_signal_type <Place signal type URI here and uncomment> ; + . +plugbase:match_output_b_cfeatures a vamp:DenseOutput ; + vamp:identifier "b_cfeatures" ; + dc:title "Conditioned B Features" ; + dc:description """Spectral features extracted from performance B, after norrmalisation and conditioning""" ; + vamp:fixed_bin_count "true" ; + vamp:unit "" ; + vamp:bin_count 84 ; +# vamp:computes_event_type <Place event type URI here and uncomment> ; +# vamp:computes_feature <Place feature attribute URI here and uncomment> ; +# vamp:computes_signal_type <Place signal type URI here and uncomment> ; + . +plugbase:match_output_overall_cost a vamp:DenseOutput ; + vamp:identifier "overall_cost" ; + dc:title "Overall Cost" ; + dc:description """Normalised overall path cost for the cheapest path""" ; + vamp:fixed_bin_count "true" ; + vamp:unit "" ; + vamp:bin_count 1 ; +# vamp:computes_event_type <Place event type URI here and uncomment> ; +# vamp:computes_feature <Place feature attribute URI here and uncomment> ; +# vamp:computes_signal_type <Place signal type URI here and uncomment> ; + .
--- a/src/FeatureExtractor.cpp Fri Feb 06 18:09:18 2015 +0000 +++ b/src/FeatureExtractor.cpp Fri Feb 06 18:41:35 2015 +0000 @@ -24,7 +24,7 @@ using namespace std; -//#define DEBUG_FEATURE_EXTRACTOR 1 +#define DEBUG_FEATURE_EXTRACTOR 1 FeatureExtractor::FeatureExtractor(Parameters parameters) : m_params(parameters) @@ -70,12 +70,15 @@ void FeatureExtractor::makeStandardFrequencyMap() { + double refFreq = m_params.referenceFrequency; double binWidth = m_params.sampleRate / m_params.fftSize; int crossoverBin = (int)(2 / (pow(2, 1/12.0) - 1)); - int crossoverMidi = lrint(log(crossoverBin*binWidth/440.0)/ + int crossoverMidi = lrint(log(crossoverBin * binWidth / refFreq)/ log(2.0) * 12 + 69); - // freq = 440 * Math.pow(2, (midi-69)/12.0) / binWidth; +#ifdef DEBUG_FEATURE_EXTRACTOR + cerr << "FeatureExtractor::makeStandardFrequencyMap: refFreq = " << refFreq << endl; +#endif int i = 0; while (i <= crossoverBin) { @@ -84,17 +87,23 @@ } while (i <= m_params.fftSize/2) { - double midi = log(i*binWidth/440.0) / log(2.0) * 12 + 69; + double midi = log(i * binWidth / refFreq) / log(2.0) * 12 + 69; if (midi > 127) midi = 127; int target = crossoverBin + lrint(midi) - crossoverMidi; if (target >= m_featureSize) target = m_featureSize - 1; m_freqMap[i++] = target; } + +#ifdef DEBUG_FEATURE_EXTRACTOR + cerr << "FeatureExtractor: crossover bin is " << crossoverBin << " for midi " + << crossoverMidi << endl; +#endif } void FeatureExtractor::makeChromaFrequencyMap() { + double refFreq = m_params.referenceFrequency; double binWidth = m_params.sampleRate / m_params.fftSize; int crossoverBin = (int)(1 / (pow(2, 1/12.0) - 1)); int i = 0; @@ -102,7 +111,7 @@ m_freqMap[i++] = 0; } while (i <= m_params.fftSize/2) { - double midi = log(i*binWidth/440.0) / log(2.0) * 12 + 69; + double midi = log(i * binWidth / refFreq) / log(2.0) * 12 + 69; m_freqMap[i++] = (lrint(midi)) % 12 + 1; } }
--- a/src/FeatureExtractor.h Fri Feb 06 18:09:18 2015 +0000 +++ b/src/FeatureExtractor.h Fri Feb 06 18:41:35 2015 +0000 @@ -48,7 +48,8 @@ Parameters(float rate_, int fftSize_) : sampleRate(rate_), useChromaFrequencyMap(false), - fftSize(fftSize_) + fftSize(fftSize_), + referenceFrequency(440.0) {} /** Sample rate of audio */ @@ -62,6 +63,9 @@ * in is already in the frequency domain, so this expresses * the size of the frame that the caller will be providing. */ int fftSize; + + /** Frequency of concert A */ + double referenceFrequency; }; /**
--- a/src/Finder.cpp Fri Feb 06 18:09:18 2015 +0000 +++ b/src/Finder.cpp Fri Feb 06 18:41:35 2015 +0000 @@ -407,6 +407,19 @@ } #endif +double +Finder::getOverallCost() +{ + int ex = m_m->getOtherFrameCount() - 1; + int ey = m_m->getFrameCount() - 1; + + if (ex < 0 || ey < 0) { + return 0; + } + + return m_m->getPathCost(ey, ex); +} + int Finder::retrievePath(bool smooth, vector<int> &pathx, vector<int> &pathy) {
--- a/src/Finder.h Fri Feb 06 18:09:18 2015 +0000 +++ b/src/Finder.h Fri Feb 06 18:41:35 2015 +0000 @@ -106,6 +106,12 @@ */ int retrievePath(bool smooth, std::vector<int> &pathx, std::vector<int> &pathy); + /** + * Get the path cost for the overall path to the end of both + * sources. + */ + double getOverallCost(); + protected: #ifdef PERFORM_ERROR_CHECKS struct ErrorPosition {
--- a/src/MatchPipeline.cpp Fri Feb 06 18:09:18 2015 +0000 +++ b/src/MatchPipeline.cpp Fri Feb 06 18:41:35 2015 +0000 @@ -20,9 +20,10 @@ MatchPipeline::MatchPipeline(FeatureExtractor::Parameters feParams, FeatureConditioner::Parameters fcParams, DistanceMetric::Parameters dParams, - Matcher::Parameters matchParams) : + Matcher::Parameters matchParams, + double secondReferenceFrequency) : m_fe1(feParams), - m_fe2(feParams), + m_fe2(paramsWithFreq(feParams, secondReferenceFrequency)), m_fc1(fcParams), m_fc2(fcParams), m_pm1(matchParams, dParams, 0), @@ -39,6 +40,14 @@ { } +FeatureExtractor::Parameters +MatchPipeline::paramsWithFreq(FeatureExtractor::Parameters params, double freq) +{ + if (freq == 0.0) return params; + params.referenceFrequency = freq; + return params; +} + void MatchPipeline::feedFrequencyDomainAudio(const float *arr1, const float *arr2) { @@ -145,5 +154,9 @@ return m_feeder.retrieveForwardPath(pathx, pathy); } +double +MatchPipeline::getOverallCost() +{ + return m_feeder.getFinder()->getOverallCost(); +} -
--- a/src/MatchPipeline.h Fri Feb 06 18:09:18 2015 +0000 +++ b/src/MatchPipeline.h Fri Feb 06 18:41:35 2015 +0000 @@ -35,11 +35,18 @@ * -> Features * -> Conditioned features * -> Matcher + * + * Only one set of FeatureExtractor::Parameters is provided; this + * contains a single reference frequency, but it's possible the + * two input streams may have different tuning frequencies. A + * separate frequency for the second input can be provided here as + * an optional parameter if needed. */ MatchPipeline(FeatureExtractor::Parameters feParams, FeatureConditioner::Parameters fcParams, DistanceMetric::Parameters dParams, - Matcher::Parameters matchParams); + Matcher::Parameters matchParams, + double secondReferenceFrequency = 0.0); ~MatchPipeline(); @@ -103,7 +110,15 @@ * See MatchFeatureFeeder::retrieveForwardPath for more details. */ void retrieveForwardPath(std::vector<int> &pathx, std::vector<int> &pathy); - + + /** + * Get the path cost for the overall path to the end of both + * sources. + * + * See Finder::getOverallCost for more details. + */ + double getOverallCost(); + private: FeatureExtractor m_fe1; FeatureExtractor m_fe2; @@ -120,6 +135,8 @@ vector<double> m_c1; vector<double> m_c2; bool aboveThreshold(const vector<double> &f); + FeatureExtractor::Parameters paramsWithFreq(FeatureExtractor::Parameters, + double); }; #endif
--- a/src/MatchVampPlugin.cpp Fri Feb 06 18:09:18 2015 +0000 +++ b/src/MatchVampPlugin.cpp Fri Feb 06 18:41:35 2015 +0000 @@ -62,6 +62,7 @@ m_defaultParams(defaultStepTime), m_feParams(inputSampleRate, m_blockSize), m_defaultFeParams(inputSampleRate, m_blockSize), + m_secondReferenceFrequency(m_defaultFeParams.referenceFrequency), m_fcParams(), m_defaultFcParams(), m_dParams(), @@ -146,19 +147,56 @@ ParameterDescriptor desc; - desc.identifier = "serialise"; - desc.name = "Serialise Plugin Invocations"; - desc.description = "Reduce potential memory load at the expense of multiprocessor performance by serialising multi-threaded plugin runs"; + desc.identifier = "freq1"; + desc.name = "Tuning frequency of first input"; + desc.description = "Tuning frequency (concert A) for the reference audio."; + desc.minValue = 220.0; + desc.maxValue = 880.0; + desc.defaultValue = (float)m_defaultFeParams.referenceFrequency; + desc.isQuantized = false; + desc.unit = "Hz"; + list.push_back(desc); + + desc.identifier = "freq2"; + desc.name = "Tuning frequency of second input"; + desc.description = "Tuning frequency (concert A) for the other audio."; + desc.minValue = 220.0; + desc.maxValue = 880.0; + desc.defaultValue = (float)m_defaultFeParams.referenceFrequency; + desc.isQuantized = false; + desc.unit = "Hz"; + list.push_back(desc); + + desc.unit = ""; + + desc.identifier = "usechroma"; + desc.name = "Feature type"; + desc.description = "Whether to use warped spectrogram or chroma frequency map"; desc.minValue = 0; desc.maxValue = 1; - desc.defaultValue = 0; + desc.defaultValue = m_defaultFeParams.useChromaFrequencyMap ? 1 : 0; + desc.isQuantized = true; + desc.quantizeStep = 1; + desc.valueNames.clear(); + desc.valueNames.push_back("Spectral"); + desc.valueNames.push_back("Chroma"); + list.push_back(desc); + + desc.valueNames.clear(); + + desc.identifier = "usespecdiff"; + desc.name = "Use feature difference"; + desc.description = "Whether to use half-wave rectified feature-to-feature difference instead of straight spectral or chroma feature"; + desc.minValue = 0; + desc.maxValue = 1; + desc.defaultValue = (int)m_defaultFcParams.order; desc.isQuantized = true; desc.quantizeStep = 1; list.push_back(desc); desc.identifier = "framenorm"; - desc.name = "Frame Normalisation"; - desc.description = "Type of normalisation to use for frequency-domain audio features"; + desc.name = "Frame normalisation"; + desc.description = "Type of normalisation to use for features"; desc.minValue = 0; desc.maxValue = 2; desc.defaultValue = (int)m_defaultFcParams.norm; @@ -166,54 +204,11 @@ desc.quantizeStep = 1; desc.valueNames.clear(); desc.valueNames.push_back("None"); - desc.valueNames.push_back("Sum To 1"); - desc.valueNames.push_back("Long-Term Average"); + desc.valueNames.push_back("Sum to 1"); + desc.valueNames.push_back("Long-term average"); list.push_back(desc); desc.valueNames.clear(); - - desc.identifier = "distnorm"; - desc.name = "Distance Normalisation"; - desc.description = "Type of normalisation to use for distance metric"; - desc.minValue = 0; - desc.maxValue = 2; - desc.defaultValue = (int)m_defaultDParams.norm; - desc.isQuantized = true; - desc.quantizeStep = 1; - desc.valueNames.clear(); - desc.valueNames.push_back("None"); - desc.valueNames.push_back("Sum of Frames"); - desc.valueNames.push_back("Log Sum of Frames"); - list.push_back(desc); - desc.valueNames.clear(); - - desc.identifier = "usespecdiff"; - desc.name = "Use Spectral Difference"; - desc.description = "Whether to use half-wave rectified spectral difference instead of straight spectrum"; - desc.minValue = 0; - desc.maxValue = 1; - desc.defaultValue = (int)m_defaultFcParams.order; - desc.isQuantized = true; - desc.quantizeStep = 1; - list.push_back(desc); - - desc.identifier = "usechroma"; - desc.name = "Use Chroma Frequency Map"; - desc.description = "Whether to use a chroma frequency map instead of the default warped spectrogram"; - desc.minValue = 0; - desc.maxValue = 1; - desc.defaultValue = m_defaultFeParams.useChromaFrequencyMap ? 1 : 0; - desc.isQuantized = true; - desc.quantizeStep = 1; - list.push_back(desc); - - desc.identifier = "silencethreshold"; - desc.name = "Silence Threshold"; - desc.description = "Total frame energy threshold below which a feature will be regarded as silent"; - desc.minValue = 0; - desc.maxValue = 1; desc.defaultValue = (float)m_defaultFcParams.silenceThreshold; - desc.isQuantized = false; - list.push_back(desc); desc.identifier = "metric"; desc.name = "Distance metric"; @@ -230,8 +225,32 @@ list.push_back(desc); desc.valueNames.clear(); + desc.identifier = "distnorm"; + desc.name = "Distance normalisation"; + desc.description = "Type of normalisation to use for distance metric"; + desc.minValue = 0; + desc.maxValue = 2; + desc.defaultValue = (int)m_defaultDParams.norm; + desc.isQuantized = true; + desc.quantizeStep = 1; + desc.valueNames.clear(); + desc.valueNames.push_back("None"); + desc.valueNames.push_back("Sum of frames"); + desc.valueNames.push_back("Log sum of frames"); + list.push_back(desc); + desc.valueNames.clear(); + + desc.identifier = "silencethreshold"; + desc.name = "Silence threshold"; + desc.description = "Total frame energy threshold below which a feature will be regarded as silent"; + desc.minValue = 0; + desc.maxValue = 0.1f; + desc.defaultValue = (float)m_defaultFcParams.silenceThreshold; + desc.isQuantized = false; + list.push_back(desc); + desc.identifier = "noise"; - desc.name = "Mix in Noise"; + desc.name = "Add noise"; 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."; desc.minValue = 0; desc.maxValue = 1; @@ -241,7 +260,7 @@ list.push_back(desc); desc.identifier = "gradientlimit"; - desc.name = "Gradient Limit"; + desc.name = "Gradient limit"; desc.description = "Limit of number of frames that will be accepted from one source without a frame from the other source being accepted"; desc.minValue = 1; desc.maxValue = 10; @@ -251,7 +270,7 @@ list.push_back(desc); desc.identifier = "zonewidth"; - desc.name = "Search Zone Width"; + desc.name = "Search zone width"; desc.description = "Width of the search zone (error margin) either side of the ongoing match position, in seconds"; desc.minValue = 1; desc.maxValue = 60; @@ -262,7 +281,7 @@ list.push_back(desc); desc.identifier = "diagonalweight"; - desc.name = "Diagonal Weight"; + desc.name = "Diagonal weight"; 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"; desc.minValue = 1.0; desc.maxValue = 2.0; @@ -272,7 +291,7 @@ list.push_back(desc); desc.identifier = "smooth"; - desc.name = "Smooth Path"; + desc.name = "Use path smoothing"; 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.)"; desc.minValue = 0; desc.maxValue = 1; @@ -282,6 +301,16 @@ desc.unit = ""; list.push_back(desc); + desc.identifier = "serialise"; + desc.name = "Serialise plugin invocations"; + desc.description = "Reduce potential memory load at the expense of multiprocessor performance by serialising multi-threaded plugin runs"; + desc.minValue = 0; + desc.maxValue = 1; + desc.defaultValue = 0; + desc.isQuantized = true; + desc.quantizeStep = 1; + list.push_back(desc); + return list; } @@ -312,6 +341,10 @@ return (int)m_dParams.metric; } else if (name == "noise") { return m_dParams.noise; + } else if (name == "freq1") { + return (float)m_feParams.referenceFrequency; + } else if (name == "freq2") { + return (float)m_secondReferenceFrequency; } return 0.0; @@ -344,6 +377,10 @@ m_dParams.metric = (DistanceMetric::Metric)(int(value + 0.1)); } else if (name == "noise") { m_dParams.noise = (DistanceMetric::NoiseAddition)(int(value + 0.1)); + } else if (name == "freq1") { + m_feParams.referenceFrequency = value; + } else if (name == "freq2") { + m_secondReferenceFrequency = value; } } @@ -365,7 +402,11 @@ m_params.hopTime = m_stepTime; m_feParams.fftSize = m_blockSize; - m_pipeline = new MatchPipeline(m_feParams, m_fcParams, m_dParams, m_params); + cerr << "creating pipeline with m_secondReferenceFrequency = " + << m_secondReferenceFrequency << endl; + m_pipeline = new MatchPipeline(m_feParams, m_fcParams, m_dParams, m_params, + m_secondReferenceFrequency); + cerr << "done" << endl; } bool @@ -532,6 +573,19 @@ m_cbFeaturesOutNo = list.size(); list.push_back(desc); + desc.identifier = "overall_cost"; + desc.name = "Overall Cost"; + desc.description = "Normalised overall path cost for the cheapest path"; + desc.unit = ""; + desc.hasFixedBinCount = true; + desc.binCount = 1; + desc.hasKnownExtents = false; + desc.isQuantized = false; + desc.sampleType = OutputDescriptor::FixedSampleRate; + desc.sampleRate = 1; + m_overallCostOutNo = list.size(); + list.push_back(desc); + return list; } @@ -610,6 +664,12 @@ std::vector<int> pathy; int len = m_pipeline->retrievePath(m_smooth, pathx, pathy); + double cost = m_pipeline->getOverallCost(); + Feature costFeature; + costFeature.hasTimestamp = false; + costFeature.values.push_back((float)cost); + returnFeatures[m_overallCostOutNo].push_back(costFeature); + int prevx = 0; int prevy = 0;
--- a/src/MatchVampPlugin.h Fri Feb 06 18:09:18 2015 +0000 +++ b/src/MatchVampPlugin.h Fri Feb 06 18:41:35 2015 +0000 @@ -78,14 +78,13 @@ bool m_smooth; int m_frameNo; - int m_lastFrameIn1; - int m_lastFrameIn2; Matcher::Parameters m_params; Matcher::Parameters m_defaultParams; FeatureExtractor::Parameters m_feParams; FeatureExtractor::Parameters m_defaultFeParams; + double m_secondReferenceFrequency; FeatureConditioner::Parameters m_fcParams; FeatureConditioner::Parameters m_defaultFcParams; @@ -102,6 +101,7 @@ mutable int m_bFeaturesOutNo; mutable int m_caFeaturesOutNo; mutable int m_cbFeaturesOutNo; + mutable int m_overallCostOutNo; #ifdef _WIN32 static HANDLE m_serialisingMutex;