Mercurial > hg > tipic

changeset 19:51eb8b1a1910

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add CRP (lacking some features)
author Chris Cannam
date Thu, 24 Sep 2015 16:35:48 +0100
parents c785eaaeac40
children 3bbdd3dada9f
files .hgsubstate Makefile.inc src/CRP.cpp src/CRP.h src/PitchFilterbank.cpp src/PitchFilterbank.h src/TipicVampPlugin.cpp src/TipicVampPlugin.h src/Types.h
diffstat 9 files changed, 175 insertions(+), 36 deletions(-) [+]
line wrap: on
line diff
--- a/.hgsubstate	Thu Sep 24 12:45:43 2015 +0100
+++ b/.hgsubstate	Thu Sep 24 16:35:48 2015 +0100
@@ -1,2 +1,2 @@
 5c47cf61c391c403a5d57ac0f7d8101ce129018b bqvec
-500416b7a651efa2a334ef0ee75fffde0d1f755d constant-q-cpp
+bc8033ed4c291e96202b32db94c9665724b2cb23 constant-q-cpp
--- a/Makefile.inc	Thu Sep 24 12:45:43 2015 +0100
+++ b/Makefile.inc	Thu Sep 24 16:35:48 2015 +0100
@@ -24,8 +24,8 @@
 
 PUBLIC_HEADERS	:=
 
-LIB_HEADERS	:= $(SRC_DIR)/delays.h $(SRC_DIR)/filter-a.h $(SRC_DIR)/filter-b.h $(SRC_DIR)/Filter.h $(SRC_DIR)/PitchFilterbank.h $(SRC_DIR)/DCT.h
-LIB_SOURCES	:= $(SRC_DIR)/Filter.cpp $(SRC_DIR)/PitchFilterbank.cpp $(SRC_DIR)/DCT.cpp
+LIB_HEADERS	:= $(SRC_DIR)/delays.h $(SRC_DIR)/filter-a.h $(SRC_DIR)/filter-b.h $(SRC_DIR)/Filter.h $(SRC_DIR)/PitchFilterbank.h $(SRC_DIR)/DCT.h $(SRC_DIR)/Types.h $(SRC_DIR)/CRP.h
+LIB_SOURCES	:= $(SRC_DIR)/Filter.cpp $(SRC_DIR)/PitchFilterbank.cpp $(SRC_DIR)/DCT.cpp $(SRC_DIR)/CRP.cpp
 LIB_OBJECTS	:= $(LIB_SOURCES:.cpp=.o)
 LIB_OBJECTS	:= $(LIB_OBJECTS:.c=.o)
 
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/CRP.cpp	Thu Sep 24 16:35:48 2015 +0100
@@ -0,0 +1,69 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+#include "CRP.h"
+
+#include "DCT.h"
+
+#include <cmath>
+#include <iostream>
+
+using namespace std;
+
+CRP::~CRP()
+{
+    delete m_dct;
+}
+
+RealBlock
+CRP::process(const RealBlock &in)
+{
+    if (in.empty()) {
+	return in;
+    }
+    if (!m_dct) {
+	m_size = in[0].size();
+	m_dct = new DCT(m_size);
+    }
+
+    int bins = 12;
+    RealBlock out;
+    RealColumn dctOut(m_size);
+
+    for (RealColumn col: in) {
+
+	RealColumn crp(bins);
+
+	if (int(col.size()) != m_size) {
+	    cerr << "ERROR: Inconsistent value count in pitch column: found "
+		 << col.size() << " where previous column(s) have had " << m_size
+		 << endl;
+	    throw std::logic_error("Inconsistent value count in pitch column");
+	}
+    
+	if (m_params.applyLogCompression) {
+	    for (auto &v: col) {
+		//!!! These numbers are parameters in the MATLAB
+		v = log10(1.0 + 1000.0 * v);
+	    }
+	}
+
+	m_dct->forward(col.data(), dctOut.data());
+
+	for (int i = 0; i < m_params.coefficientsToDrop; ++i) {
+	    dctOut[i] = 0.0;
+	}
+
+	m_dct->inverse(dctOut.data(), col.data());
+
+	for (int i = 0; i < m_size; ++i) {
+	    crp[i % bins] += col[i];
+	}
+
+	//!!! normalise as in normalizeFeature.m
+	
+	out.push_back(crp);
+    }
+
+    return out;
+}
+
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/CRP.h	Thu Sep 24 16:35:48 2015 +0100
@@ -0,0 +1,33 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+#ifndef CRP_H
+#define CRP_H
+
+#include "Types.h"
+
+class DCT;
+
+//!!! Downsampling/temporal smoothing not yet implemented
+
+class CRP
+{
+public:
+    struct Parameters {
+	int coefficientsToDrop;
+	bool applyLogCompression;
+	Parameters() : coefficientsToDrop(54), applyLogCompression(true) { }
+    };
+	
+    CRP(Parameters params) : m_params(params), m_size(0), m_dct(0) { }
+    ~CRP();
+
+    RealBlock process(const RealBlock &in);
+    
+private:
+    Parameters m_params;
+    int m_size;
+    DCT *m_dct;
+};
+
+#endif
+
--- a/src/PitchFilterbank.cpp	Thu Sep 24 12:45:43 2015 +0100
+++ b/src/PitchFilterbank.cpp	Thu Sep 24 16:35:48 2015 +0100
@@ -63,15 +63,6 @@
 
     int getSampleRate() const { return m_sampleRate; }
     float getTuningFrequency() const { return m_tuningFrequency; }
-    
-    /// A series of real-valued samples ordered in time.
-    typedef vector<double> RealSequence;
-
-    /// A series of real-valued samples ordered by bin (frequency or similar).
-    typedef vector<double> RealColumn;
-
-    /// A matrix of real-valued samples, indexed by time then bin number.
-    typedef vector<RealColumn> RealBlock;
 
     RealBlock process(const RealSequence &in) {
 
@@ -277,13 +268,13 @@
     m_d = new D(rate, freq);
 }
 
-PitchFilterbank::RealBlock
+RealBlock
 PitchFilterbank::process(const RealSequence &in)
 {
     return m_d->process(in);
 }
 
-PitchFilterbank::RealBlock
+RealBlock
 PitchFilterbank::getRemainingOutput()
 {
     return m_d->getRemainingOutput();
--- a/src/PitchFilterbank.h	Thu Sep 24 12:45:43 2015 +0100
+++ b/src/PitchFilterbank.h	Thu Sep 24 16:35:48 2015 +0100
@@ -3,7 +3,7 @@
 #ifndef PITCH_FILTERBANK_H
 #define PITCH_FILTERBANK_H
 
-#include <vector>
+#include "Types.h"
 
 class PitchFilterbank
 {
@@ -11,15 +11,6 @@
     PitchFilterbank(int sampleRate, float tuningFrequency);
     ~PitchFilterbank();
 
-    /// A series of real-valued samples ordered in time.
-    typedef std::vector<double> RealSequence;
-
-    /// A series of real-valued samples ordered by bin (frequency or similar).
-    typedef std::vector<double> RealColumn;
-
-    /// A matrix of real-valued samples, indexed by time then bin number.
-    typedef std::vector<RealColumn> RealBlock;
-
     void reset();
     
     RealBlock process(const RealSequence &);
--- a/src/TipicVampPlugin.cpp	Thu Sep 24 12:45:43 2015 +0100
+++ b/src/TipicVampPlugin.cpp	Thu Sep 24 16:35:48 2015 +0100
@@ -1,6 +1,9 @@
 
 #include "TipicVampPlugin.h"
 
+#include "PitchFilterbank.h"
+#include "CRP.h"
+
 #include <bqvec/Range.h>
 #include <bqvec/VectorOps.h>
 
@@ -18,7 +21,9 @@
     m_blockSize(0),
     m_tuningFrequency(defaultTuningFrequency),
     m_filterbank(0),
-    m_pitchOutputNo(-1)
+    m_crp(0),
+    m_pitchOutputNo(-1),
+    m_crpOutputNo(-1)
 {
 }
 
@@ -177,21 +182,44 @@
     m_pitchOutputNo = list.size();
     list.push_back(d);
 
+    d.identifier = "crp";
+    d.name = "Chroma DCT-Reduced Log Pitch Features";
+    d.description = "";
+    d.unit = "";
+    d.hasFixedBinCount = true;
+    d.binCount = 12; //!!! make parameter in CRP
+    d.hasKnownExtents = false;
+    d.isQuantized = false;
+    d.sampleType = OutputDescriptor::FixedSampleRate;
+    d.sampleRate = 22050 / 2205; //!!! get block size & hop from filterbank
+    d.hasDuration = false;
+    m_crpOutputNo = list.size();
+    list.push_back(d);
+
     return list;
 }
 
 bool
 Tipic::initialise(size_t channels, size_t stepSize, size_t blockSize)
 {
+    if (m_inputSampleRate > 192000) {
+	cerr << "ERROR: Tipic::initialise: Max sample rate 192000 exceeded "
+	     << "(requested rate = " << m_inputSampleRate << ")" << endl;
+	return false;
+    }
+    
     if (m_pitchOutputNo < 0) {
 	// getOutputDescriptors has never been called, it sets up the
 	// outputNo members
 	(void)getOutputDescriptors();
     }
+    if (m_pitchOutputNo < 0 || m_crpOutputNo < 0) {
+	throw std::logic_error("setup went wrong");
+    }
     
     if (channels < getMinChannelCount() ||
 	channels > getMaxChannelCount()) {
-	cerr << "ERROR: initialise: wrong number of channels supplied (only 1 supported)" << endl;
+	cerr << "ERROR: Tipic::initialise: wrong number of channels supplied (only 1 supported)" << endl;
 	return false;
     }
 
@@ -213,6 +241,7 @@
 {
     if (!m_filterbank) {
 	m_filterbank = new PitchFilterbank(m_inputSampleRate, m_tuningFrequency);
+	m_crp = new CRP({});
     }
     m_filterbank->reset();
 }
@@ -220,36 +249,39 @@
 Tipic::FeatureSet
 Tipic::process(const float *const *inputBuffers, Vamp::RealTime timestamp)
 {
-    PitchFilterbank::RealSequence in;
+    RealSequence in;
     in.resize(m_blockSize);
     v_convert(in.data(), inputBuffers[0], m_blockSize);
     
-    PitchFilterbank::RealBlock pitchFiltered = m_filterbank->process(in);
+    RealBlock pitchFiltered = m_filterbank->process(in);
+    RealBlock crpReduced = m_crp->process(pitchFiltered);
 
     FeatureSet fs;
-    addPitchFeatures(fs, pitchFiltered);
+    addFeatures(fs, m_pitchOutputNo, pitchFiltered);
+    addFeatures(fs, m_crpOutputNo, crpReduced);
     return fs;
 }
 
 Tipic::FeatureSet
 Tipic::getRemainingFeatures()
 {
-    PitchFilterbank::RealBlock pitchFiltered = m_filterbank->getRemainingOutput();
+    RealBlock pitchFiltered = m_filterbank->getRemainingOutput();
+    RealBlock crpReduced = m_crp->process(pitchFiltered);
 
     FeatureSet fs;
-    addPitchFeatures(fs, pitchFiltered);
+    addFeatures(fs, m_pitchOutputNo, pitchFiltered);
+    addFeatures(fs, m_crpOutputNo, crpReduced);
     return fs;
 }
 
 void
-Tipic::addPitchFeatures(FeatureSet &fs, const PitchFilterbank::RealBlock &block)
+Tipic::addFeatures(FeatureSet &fs, int outputNo, const RealBlock &block)
 {
     for (int i = 0; in_range_for(block, i); ++i) {
 	Feature f;
 	int h = block[i].size();
 	f.values.resize(h);
 	v_convert(f.values.data(), block[i].data(), h);
-	fs[m_pitchOutputNo].push_back(f);
+	fs[outputNo].push_back(f);
     }
 }
-
--- a/src/TipicVampPlugin.h	Thu Sep 24 12:45:43 2015 +0100
+++ b/src/TipicVampPlugin.h	Thu Sep 24 16:35:48 2015 +0100
@@ -3,7 +3,10 @@
 
 #include <vamp-sdk/Plugin.h>
 
-#include "PitchFilterbank.h"
+#include "Types.h"
+
+class PitchFilterbank;
+class CRP;
 
 using std::string;
 
@@ -49,9 +52,11 @@
     int m_blockSize;
     float m_tuningFrequency;
     PitchFilterbank *m_filterbank;
+    CRP *m_crp;
     mutable int m_pitchOutputNo;
+    mutable int m_crpOutputNo;
 
-    void addPitchFeatures(FeatureSet &, const PitchFilterbank::RealBlock &);
+    void addFeatures(FeatureSet &, int outputNo, const RealBlock &);
 };
 
 
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/Types.h	Thu Sep 24 16:35:48 2015 +0100
@@ -0,0 +1,18 @@
+/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
+
+#ifndef TIPIC_TYPES_H
+#define TIPIC_TYPES_H
+
+#include <vector>
+
+/// A series of real-valued samples ordered in time.
+typedef std::vector<double> RealSequence;
+
+/// A series of real-valued samples ordered by bin (frequency or similar).
+typedef std::vector<double> RealColumn;
+
+/// A matrix of real-valued samples, indexed by time then bin number.
+typedef std::vector<RealColumn> RealBlock;
+
+#endif
+