Chris@31: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
Chris@31: 
Chris@31: /*
Chris@31:   Silvet
Chris@31: 
Chris@31:   A Vamp plugin for note transcription.
Chris@31:   Centre for Digital Music, Queen Mary University of London.
Chris@31:     
Chris@31:   This program is free software; you can redistribute it and/or
Chris@31:   modify it under the terms of the GNU General Public License as
Chris@31:   published by the Free Software Foundation; either version 2 of the
Chris@31:   License, or (at your option) any later version.  See the file
Chris@31:   COPYING included with this distribution for more information.
Chris@31: */
Chris@31: 
Chris@31: #include "Silvet.h"
Chris@31: 
Chris@31: #include "data/include/templates.h"
Chris@31: 
Chris@32: #include "maths/MedianFilter.h"
Chris@31: #include "dsp/rateconversion/Resampler.h"
Chris@31: 
Chris@32: #include "constant-q-cpp/cpp-qm-dsp/CQInterpolated.h"
Chris@31: 
Chris@31: #include <vector>
Chris@31: 
Chris@32: #include <cstdio>
Chris@32: 
Chris@31: using std::vector;
Chris@31: using std::cerr;
Chris@31: using std::endl;
Chris@31: 
Chris@31: static int processingSampleRate = 44100;
Chris@31: static int processingBPO = 60;
Chris@32: static int processingHeight = 545;
Chris@31: 
Chris@31: Silvet::Silvet(float inputSampleRate) :
Chris@31:     Plugin(inputSampleRate),
Chris@31:     m_resampler(0),
Chris@31:     m_cq(0)
Chris@31: {
Chris@31: }
Chris@31: 
Chris@31: Silvet::~Silvet()
Chris@31: {
Chris@31:     delete m_resampler;
Chris@31:     delete m_cq;
Chris@32:     for (int i = 0; i < (int)m_filterA.size(); ++i) {
Chris@32:         delete m_filterA[i];
Chris@32:         delete m_filterB[i];
Chris@32:     }
Chris@31: }
Chris@31: 
Chris@31: string
Chris@31: Silvet::getIdentifier() const
Chris@31: {
Chris@31:     return "silvet";
Chris@31: }
Chris@31: 
Chris@31: string
Chris@31: Silvet::getName() const
Chris@31: {
Chris@31:     return "Silvet Note Transcription";
Chris@31: }
Chris@31: 
Chris@31: string
Chris@31: Silvet::getDescription() const
Chris@31: {
Chris@31:     // Return something helpful here!
Chris@31:     return "";
Chris@31: }
Chris@31: 
Chris@31: string
Chris@31: Silvet::getMaker() const
Chris@31: {
Chris@31:     // Your name here
Chris@31:     return "";
Chris@31: }
Chris@31: 
Chris@31: int
Chris@31: Silvet::getPluginVersion() const
Chris@31: {
Chris@31:     return 1;
Chris@31: }
Chris@31: 
Chris@31: string
Chris@31: Silvet::getCopyright() const
Chris@31: {
Chris@31:     // This function is not ideally named.  It does not necessarily
Chris@31:     // need to say who made the plugin -- getMaker does that -- but it
Chris@31:     // should indicate the terms under which it is distributed.  For
Chris@31:     // example, "Copyright (year). All Rights Reserved", or "GPL"
Chris@31:     return "";
Chris@31: }
Chris@31: 
Chris@31: Silvet::InputDomain
Chris@31: Silvet::getInputDomain() const
Chris@31: {
Chris@31:     return TimeDomain;
Chris@31: }
Chris@31: 
Chris@31: size_t
Chris@31: Silvet::getPreferredBlockSize() const
Chris@31: {
Chris@31:     return 0;
Chris@31: }
Chris@31: 
Chris@31: size_t 
Chris@31: Silvet::getPreferredStepSize() const
Chris@31: {
Chris@31:     return 0;
Chris@31: }
Chris@31: 
Chris@31: size_t
Chris@31: Silvet::getMinChannelCount() const
Chris@31: {
Chris@31:     return 1;
Chris@31: }
Chris@31: 
Chris@31: size_t
Chris@31: Silvet::getMaxChannelCount() const
Chris@31: {
Chris@31:     return 1;
Chris@31: }
Chris@31: 
Chris@31: Silvet::ParameterList
Chris@31: Silvet::getParameterDescriptors() const
Chris@31: {
Chris@31:     ParameterList list;
Chris@31:     return list;
Chris@31: }
Chris@31: 
Chris@31: float
Chris@31: Silvet::getParameter(string identifier) const
Chris@31: {
Chris@31:     return 0;
Chris@31: }
Chris@31: 
Chris@31: void
Chris@31: Silvet::setParameter(string identifier, float value) 
Chris@31: {
Chris@31: }
Chris@31: 
Chris@31: Silvet::ProgramList
Chris@31: Silvet::getPrograms() const
Chris@31: {
Chris@31:     ProgramList list;
Chris@31:     return list;
Chris@31: }
Chris@31: 
Chris@31: string
Chris@31: Silvet::getCurrentProgram() const
Chris@31: {
Chris@31:     return ""; 
Chris@31: }
Chris@31: 
Chris@31: void
Chris@31: Silvet::selectProgram(string name)
Chris@31: {
Chris@31: }
Chris@31: 
Chris@31: Silvet::OutputList
Chris@31: Silvet::getOutputDescriptors() const
Chris@31: {
Chris@31:     OutputList list;
Chris@31: 
Chris@31:     OutputDescriptor d;
Chris@31:     d.identifier = "transcription";
Chris@31:     d.name = "Transcription";
Chris@31:     d.description = ""; //!!!
Chris@31:     d.unit = "Hz";
Chris@31:     d.hasFixedBinCount = true;
Chris@31:     d.binCount = 2;
Chris@31:     d.binNames.push_back("Frequency");
Chris@31:     d.binNames.push_back("Velocity");
Chris@31:     d.hasKnownExtents = false;
Chris@31:     d.isQuantized = false;
Chris@31:     d.sampleType = OutputDescriptor::VariableSampleRate;
Chris@32:     d.sampleRate = m_inputSampleRate / (m_cq ? m_cq->getColumnHop() : 256);
Chris@31:     d.hasDuration = true;
Chris@32:     m_notesOutputNo = list.size();
Chris@32:     list.push_back(d);
Chris@32: 
Chris@32:     d.identifier = "inputgrid";
Chris@32:     d.name = "Filtered time-frequency grid";
Chris@32:     d.description = "The pre-processed constant-Q time-frequency distribution used as input to the PLCA step";
Chris@32:     d.unit = "";
Chris@32:     d.hasFixedBinCount = true;
Chris@32:     d.binCount = processingHeight;
Chris@32:     d.binNames.clear();
Chris@32:     if (m_cq) {
Chris@32:         char name[20];
Chris@32:         for (int i = 0; i < processingHeight; ++i) {
Chris@32:             float freq = m_cq->getBinFrequency(i + 55);
Chris@32:             sprintf(name, "%.1f Hz", freq);
Chris@32:             d.binNames.push_back(name);
Chris@32:         }
Chris@32:     }
Chris@32:     d.hasKnownExtents = false;
Chris@32:     d.isQuantized = false;
Chris@32:     d.sampleType = OutputDescriptor::FixedSampleRate;
Chris@32:     d.sampleRate = 25;
Chris@32:     d.hasDuration = false;
Chris@32:     m_cqOutputNo = list.size();
Chris@31:     list.push_back(d);
Chris@31: 
Chris@31:     return list;
Chris@31: }
Chris@31: 
Chris@31: bool
Chris@31: Silvet::initialise(size_t channels, size_t stepSize, size_t blockSize)
Chris@31: {
Chris@31:     if (channels < getMinChannelCount() ||
Chris@31: 	channels > getMaxChannelCount()) return false;
Chris@31: 
Chris@31:     if (stepSize != blockSize) {
Chris@31: 	cerr << "Silvet::initialise: Step size must be the same as block size ("
Chris@31: 	     << stepSize << " != " << blockSize << ")" << endl;
Chris@31: 	return false;
Chris@31:     }
Chris@31: 
Chris@31:     m_blockSize = blockSize;
Chris@31: 
Chris@31:     reset();
Chris@31: 
Chris@31:     return true;
Chris@31: }
Chris@31: 
Chris@31: void
Chris@31: Silvet::reset()
Chris@31: {
Chris@31:     delete m_resampler;
Chris@31:     delete m_cq;
Chris@31: 
Chris@31:     if (m_inputSampleRate != processingSampleRate) {
Chris@31: 	m_resampler = new Resampler(m_inputSampleRate, processingSampleRate);
Chris@31:     } else {
Chris@31: 	m_resampler = 0;
Chris@31:     }
Chris@31: 
Chris@32:     m_cq = new CQInterpolated
Chris@32: 	(processingSampleRate, 27.5, processingSampleRate / 3, processingBPO,
Chris@32:          CQInterpolated::Linear);
Chris@31: 
Chris@32:     for (int i = 0; i < (int)m_filterA.size(); ++i) {
Chris@32:         delete m_filterA[i];
Chris@32:         delete m_filterB[i];
Chris@32:     }
Chris@32:     m_filterA.clear();
Chris@32:     m_filterB.clear();
Chris@32:     for (int i = 0; i < processingHeight; ++i) {
Chris@32:         m_filterA.push_back(new MedianFilter<double>(40));
Chris@32:         m_filterB.push_back(new MedianFilter<double>(40));
Chris@32:     }
Chris@32:     m_columnCount = 0;
Chris@32:     m_reducedColumnCount = 0;
Chris@31: }
Chris@31: 
Chris@31: Silvet::FeatureSet
Chris@31: Silvet::process(const float *const *inputBuffers, Vamp::RealTime timestamp)
Chris@31: {
Chris@31:     vector<double> data;
Chris@31:     for (int i = 0; i < m_blockSize; ++i) data.push_back(inputBuffers[0][i]);
Chris@31: 
Chris@31:     if (m_resampler) {
Chris@31: 	data = m_resampler->process(data.data(), data.size());
Chris@31:     }
Chris@31: 
Chris@32:     Grid cqout = m_cq->process(data);
Chris@32:     Grid filtered = preProcess(cqout);
Chris@31: 
Chris@32:     FeatureSet fs;
Chris@32: 
Chris@32:     for (int i = 0; i < (int)filtered.size(); ++i) {
Chris@32:         Feature f;
Chris@32:         for (int j = 0; j < processingHeight; ++j) {
Chris@32:             f.values.push_back(float(filtered[i][j]));
Chris@32:         }
Chris@32:         fs[m_cqOutputNo].push_back(f);
Chris@32:     }
Chris@32: 
Chris@32:     return fs;
Chris@31: }
Chris@31: 
Chris@31: Silvet::FeatureSet
Chris@31: Silvet::getRemainingFeatures()
Chris@31: {
Chris@31: 
Chris@31:     return FeatureSet();
Chris@31: }
Chris@31: 
Chris@32: Silvet::Grid
Chris@32: Silvet::preProcess(const Grid &in)
Chris@32: {
Chris@32:     int width = in.size();
Chris@32: 
Chris@32:     // reduce to 100 columns per second, or one column every 441 samples
Chris@32: 
Chris@32:     int spacing = processingSampleRate / 100;
Chris@32: 
Chris@32:     Grid out;
Chris@32: 
Chris@32:     for (int i = 0; i < width; ++i) {
Chris@32: 
Chris@32:         int prevSampleNo = (m_columnCount - 1) * m_cq->getColumnHop();
Chris@32:         int sampleNo = m_columnCount * m_cq->getColumnHop();
Chris@32: 
Chris@32:         bool select = (sampleNo / spacing != prevSampleNo / spacing);
Chris@32: 
Chris@32:         if (select) {
Chris@32:             vector<double> inCol = in[i];
Chris@32:             vector<double> outCol(processingHeight);
Chris@32: 
Chris@32:             // we reverse the column as we go (the CQ output is
Chris@32:             // "upside-down", with high frequencies at the start of
Chris@32:             // each column, and we want it the other way around) and
Chris@32:             // then ignore the first 55 (lowest-frequency) bins,
Chris@32:             // giving us 545 bins instead of 600
Chris@32: 
Chris@32:             for (int j = 0; j < processingHeight; ++j) {
Chris@32: 
Chris@32:                 int ix = inCol.size() - j - 55;
Chris@32: 
Chris@32:                 double val = inCol[ix];
Chris@32:                 m_filterA[j]->push(val);
Chris@32: 
Chris@32:                 double a = m_filterA[j]->get();
Chris@32:                 m_filterB[j]->push(std::min(a, val));
Chris@32: 
Chris@32:                 double filtered = m_filterB[j]->get();
Chris@32:                 outCol[j] = filtered;
Chris@32:             }
Chris@32: 
Chris@32:             // then we only use every fourth filtered column, for 25
Chris@32:             // columns per second in the eventual grid
Chris@32: 
Chris@32:             if (m_reducedColumnCount % 4 == 0) {
Chris@32:                 out.push_back(outCol);
Chris@32:             }
Chris@32: 
Chris@32:             ++m_reducedColumnCount;
Chris@32:         }
Chris@32: 
Chris@32:         ++m_columnCount;
Chris@32:     }
Chris@32: 
Chris@32:     return out;
Chris@32: }
Chris@32: