Chris@152: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
Chris@152: 
Chris@152: /*
Chris@152:     Sonic Visualiser
Chris@152:     An audio file viewer and annotation editor.
Chris@152:     Centre for Digital Music, Queen Mary, University of London.
Chris@152:     This file copyright 2006 Chris Cannam.
Chris@152:     
Chris@152:     This program is free software; you can redistribute it and/or
Chris@152:     modify it under the terms of the GNU General Public License as
Chris@152:     published by the Free Software Foundation; either version 2 of the
Chris@152:     License, or (at your option) any later version.  See the file
Chris@152:     COPYING included with this distribution for more information.
Chris@152: */
Chris@152: 
Chris@152: #include "FFTModel.h"
Chris@152: #include "DenseTimeValueModel.h"
Chris@152: 
Chris@183: #include "base/Profiler.h"
Chris@275: #include "base/Pitch.h"
Chris@183: 
Chris@152: #include <cassert>
Chris@152: 
Chris@152: FFTModel::FFTModel(const DenseTimeValueModel *model,
Chris@152:                    int channel,
Chris@152:                    WindowType windowType,
Chris@152:                    size_t windowSize,
Chris@152:                    size_t windowIncrement,
Chris@152:                    size_t fftSize,
Chris@152:                    bool polar,
Chris@152:                    size_t fillFromColumn) :
Chris@152:     //!!! ZoomConstraint!
Chris@152:     m_server(0),
Chris@152:     m_xshift(0),
Chris@152:     m_yshift(0)
Chris@152: {
Chris@152:     m_server = FFTDataServer::getFuzzyInstance(model,
Chris@152:                                                channel,
Chris@152:                                                windowType,
Chris@152:                                                windowSize,
Chris@152:                                                windowIncrement,
Chris@152:                                                fftSize,
Chris@152:                                                polar,
Chris@152:                                                fillFromColumn);
Chris@152: 
Chris@200:     if (!m_server) return; // caller should check isOK()
Chris@200: 
Chris@152:     size_t xratio = windowIncrement / m_server->getWindowIncrement();
Chris@152:     size_t yratio = m_server->getFFTSize() / fftSize;
Chris@152: 
Chris@152:     while (xratio > 1) {
Chris@152:         if (xratio & 0x1) {
Chris@152:             std::cerr << "ERROR: FFTModel: Window increment ratio "
Chris@152:                       << windowIncrement << " / "
Chris@152:                       << m_server->getWindowIncrement()
Chris@152:                       << " must be a power of two" << std::endl;
Chris@152:             assert(!(xratio & 0x1));
Chris@152:         }
Chris@152:         ++m_xshift;
Chris@152:         xratio >>= 1;
Chris@152:     }
Chris@152: 
Chris@152:     while (yratio > 1) {
Chris@152:         if (yratio & 0x1) {
Chris@152:             std::cerr << "ERROR: FFTModel: FFT size ratio "
Chris@152:                       << m_server->getFFTSize() << " / " << fftSize
Chris@152:                       << " must be a power of two" << std::endl;
Chris@152:             assert(!(yratio & 0x1));
Chris@152:         }
Chris@152:         ++m_yshift;
Chris@152:         yratio >>= 1;
Chris@152:     }
Chris@152: }
Chris@152: 
Chris@152: FFTModel::~FFTModel()
Chris@152: {
Chris@200:     if (m_server) FFTDataServer::releaseInstance(m_server);
Chris@152: }
Chris@152: 
Chris@152: size_t
Chris@152: FFTModel::getSampleRate() const
Chris@152: {
Chris@152:     return isOK() ? m_server->getModel()->getSampleRate() : 0;
Chris@152: }
Chris@152: 
Chris@152: void
Chris@182: FFTModel::getColumn(size_t x, Column &result) const
Chris@152: {
Chris@183:     Profiler profiler("FFTModel::getColumn", false);
Chris@183: 
Chris@152:     result.clear();
Chris@152:     size_t height(getHeight());
Chris@152:     for (size_t y = 0; y < height; ++y) {
Chris@152:         result.push_back(const_cast<FFTModel *>(this)->getMagnitudeAt(x, y));
Chris@152:     }
Chris@152: }
Chris@152: 
Chris@152: QString
Chris@152: FFTModel::getBinName(size_t n) const
Chris@152: {
Chris@152:     size_t sr = getSampleRate();
Chris@152:     if (!sr) return "";
Chris@204:     QString name = tr("%1 Hz").arg((n * sr) / ((getHeight()-1) * 2));
Chris@152:     return name;
Chris@152: }
Chris@152: 
Chris@275: bool
Chris@275: FFTModel::estimateStableFrequency(size_t x, size_t y, float &frequency)
Chris@275: {
Chris@275:     if (!isOK()) return false;
Chris@275: 
Chris@275:     size_t sampleRate = m_server->getModel()->getSampleRate();
Chris@275: 
Chris@275:     size_t fftSize = m_server->getFFTSize() >> m_yshift;
Chris@275:     frequency = (float(y) * sampleRate) / fftSize;
Chris@275: 
Chris@275:     if (x+1 >= getWidth()) return false;
Chris@275: 
Chris@275:     // At frequency f, a phase shift of 2pi (one cycle) happens in 1/f sec.
Chris@275:     // At hopsize h and sample rate sr, one hop happens in h/sr sec.
Chris@275:     // At window size w, for bin b, f is b*sr/w.
Chris@275:     // thus 2pi phase shift happens in w/(b*sr) sec.
Chris@275:     // We need to know what phase shift we expect from h/sr sec.
Chris@275:     // -> 2pi * ((h/sr) / (w/(b*sr)))
Chris@275:     //  = 2pi * ((h * b * sr) / (w * sr))
Chris@275:     //  = 2pi * (h * b) / w.
Chris@275: 
Chris@275:     float oldPhase = getPhaseAt(x, y);
Chris@275:     float newPhase = getPhaseAt(x+1, y);
Chris@275: 
Chris@275:     size_t incr = getResolution();
Chris@275: 
Chris@275:     float expectedPhase = oldPhase + (2.0 * M_PI * y * incr) / fftSize;
Chris@275: 
Chris@275:     float phaseError = princargf(newPhase - expectedPhase);
Chris@275: 
Chris@275: //    bool stable = (fabsf(phaseError) < (1.1f * (m_windowIncrement * M_PI) / m_fftSize));
Chris@275: 
Chris@275:     // The new frequency estimate based on the phase error resulting
Chris@275:     // from assuming the "native" frequency of this bin
Chris@275: 
Chris@275:     frequency =
Chris@275:         (sampleRate * (expectedPhase + phaseError - oldPhase)) /
Chris@275:         (2 * M_PI * incr);
Chris@275: 
Chris@275:     return true;
Chris@275: }
Chris@275: 
Chris@275: FFTModel::PeakLocationSet
Chris@275: FFTModel::getPeaks(PeakPickType type, size_t x, size_t ymin, size_t ymax)
Chris@275: {
Chris@275:     FFTModel::PeakLocationSet peaks;
Chris@275:     if (!isOK()) return peaks;
Chris@275: 
Chris@275:     if (ymax == 0 || ymax > getHeight() - 1) {
Chris@275:         ymax = getHeight() - 1;
Chris@275:     }
Chris@275: 
Chris@275:     Column values;
Chris@275: 
Chris@275:     if (type == AllPeaks) {
Chris@275:         for (size_t y = ymin; y <= ymax; ++y) {
Chris@275:             values.push_back(getMagnitudeAt(x, y));
Chris@275:         }
Chris@275:         size_t i = 0;
Chris@275:         for (size_t bin = ymin; bin <= ymax; ++bin) {
Chris@275:             if ((i == 0 || values[i] > values[i-1]) &&
Chris@275:                 (i == values.size()-1 || values[i] >= values[i+1])) {
Chris@275:                 peaks.insert(bin);
Chris@275:             }
Chris@275:             ++i;
Chris@275:         }
Chris@275:         return peaks;
Chris@275:     }
Chris@275: 
Chris@275:     getColumn(x, values);
Chris@275: 
Chris@275:     // For peak picking we use a moving median window, picking the
Chris@275:     // highest value within each continuous region of values that
Chris@275:     // exceed the median.  For pitch adaptivity, we adjust the window
Chris@275:     // size to a roughly constant pitch range (about four tones).
Chris@275: 
Chris@275:     size_t sampleRate = getSampleRate();
Chris@275: 
Chris@275:     std::deque<float> window;
Chris@275:     std::vector<size_t> inrange;
Chris@280:     float dist = 0.5;
Chris@280:     size_t medianWinSize = getPeakPickWindowSize(type, sampleRate, ymin, dist);
Chris@275:     size_t halfWin = medianWinSize/2;
Chris@275: 
Chris@275:     size_t binmin;
Chris@275:     if (ymin > halfWin) binmin = ymin - halfWin;
Chris@275:     else binmin = 0;
Chris@275: 
Chris@275:     size_t binmax;
Chris@275:     if (ymax + halfWin < values.size()) binmax = ymax + halfWin;
Chris@275:     else binmax = values.size()-1;
Chris@275: 
Chris@275:     for (size_t bin = binmin; bin <= binmax; ++bin) {
Chris@275: 
Chris@275:         float value = values[bin];
Chris@275: 
Chris@275:         window.push_back(value);
Chris@275: 
Chris@280:         // so-called median will actually be the dist*100'th percentile
Chris@280:         medianWinSize = getPeakPickWindowSize(type, sampleRate, bin, dist);
Chris@275:         halfWin = medianWinSize/2;
Chris@275: 
Chris@275:         while (window.size() > medianWinSize) window.pop_front();
Chris@275: 
Chris@275:         if (type == MajorPitchAdaptivePeaks) {
Chris@275:             if (ymax + halfWin < values.size()) binmax = ymax + halfWin;
Chris@275:             else binmax = values.size()-1;
Chris@275:         }
Chris@275: 
Chris@275:         std::deque<float> sorted(window);
Chris@275:         std::sort(sorted.begin(), sorted.end());
Chris@280:         float median = sorted[int(sorted.size() * dist)];
Chris@275: 
Chris@275:         if (value > median) {
Chris@275:             inrange.push_back(bin);
Chris@275:         }
Chris@275: 
Chris@275:         if (value <= median || bin+1 == values.size()) {
Chris@275:             size_t peakbin = 0;
Chris@275:             float peakval = 0.f;
Chris@275:             if (!inrange.empty()) {
Chris@275:                 for (size_t i = 0; i < inrange.size(); ++i) {
Chris@275:                     if (i == 0 || values[inrange[i]] > peakval) {
Chris@275:                         peakval = values[inrange[i]];
Chris@275:                         peakbin = inrange[i];
Chris@275:                     }
Chris@275:                 }
Chris@275:                 inrange.clear();
Chris@275:                 if (peakbin >= ymin && peakbin <= ymax) {
Chris@275:                     peaks.insert(peakbin);
Chris@275:                 }
Chris@275:             }
Chris@275:         }
Chris@275:     }
Chris@275: 
Chris@275:     return peaks;
Chris@275: }
Chris@275: 
Chris@275: size_t
Chris@280: FFTModel::getPeakPickWindowSize(PeakPickType type, size_t sampleRate,
Chris@280:                                 size_t bin, float &percentile) const
Chris@275: {
Chris@280:     percentile = 0.5;
Chris@275:     if (type == MajorPeaks) return 10;
Chris@275:     if (bin == 0) return 3;
Chris@280: 
Chris@275:     size_t fftSize = m_server->getFFTSize() >> m_yshift;
Chris@275:     float binfreq = (sampleRate * bin) / fftSize;
Chris@275:     float hifreq = Pitch::getFrequencyForPitch(73, 0, binfreq);
Chris@280: 
Chris@275:     int hibin = lrintf((hifreq * fftSize) / sampleRate);
Chris@275:     int medianWinSize = hibin - bin;
Chris@275:     if (medianWinSize < 3) medianWinSize = 3;
Chris@280: 
Chris@280:     percentile = 0.5 + (binfreq / sampleRate);
Chris@280: 
Chris@275:     return medianWinSize;
Chris@275: }
Chris@275: 
Chris@275: FFTModel::PeakSet
Chris@275: FFTModel::getPeakFrequencies(PeakPickType type, size_t x,
Chris@275:                              size_t ymin, size_t ymax)
Chris@275: {
Chris@275:     PeakSet peaks;
Chris@275:     if (!isOK()) return peaks;
Chris@275:     PeakLocationSet locations = getPeaks(type, x, ymin, ymax);
Chris@275: 
Chris@275:     size_t sampleRate = getSampleRate();
Chris@275:     size_t fftSize = m_server->getFFTSize() >> m_yshift;
Chris@275:     size_t incr = getResolution();
Chris@275: 
Chris@275:     // This duplicates some of the work of estimateStableFrequency to
Chris@275:     // allow us to retrieve the phases in two separate vertical
Chris@275:     // columns, instead of jumping back and forth between columns x and
Chris@275:     // x+1, which may be significantly slower if re-seeking is needed
Chris@275: 
Chris@275:     std::vector<float> phases;
Chris@275:     for (PeakLocationSet::iterator i = locations.begin();
Chris@275:          i != locations.end(); ++i) {
Chris@275:         phases.push_back(getPhaseAt(x, *i));
Chris@275:     }
Chris@275: 
Chris@275:     size_t phaseIndex = 0;
Chris@275:     for (PeakLocationSet::iterator i = locations.begin();
Chris@275:          i != locations.end(); ++i) {
Chris@275:         float oldPhase = phases[phaseIndex];
Chris@275:         float newPhase = getPhaseAt(x+1, *i);
Chris@275:         float expectedPhase = oldPhase + (2.0 * M_PI * *i * incr) / fftSize;
Chris@275:         float phaseError = princargf(newPhase - expectedPhase);
Chris@275:         float frequency =
Chris@275:             (sampleRate * (expectedPhase + phaseError - oldPhase))
Chris@275:             / (2 * M_PI * incr);
Chris@275: //        bool stable = (fabsf(phaseError) < (1.1f * (incr * M_PI) / fftSize));
Chris@275: //        if (stable)
Chris@275:         peaks[*i] = frequency;
Chris@275:         ++phaseIndex;
Chris@275:     }
Chris@275: 
Chris@275:     return peaks;
Chris@275: }
Chris@275: 
Chris@152: Model *
Chris@152: FFTModel::clone() const
Chris@152: {
Chris@152:     return new FFTModel(*this);
Chris@152: }
Chris@152: 
Chris@152: FFTModel::FFTModel(const FFTModel &model) :
Chris@152:     DenseThreeDimensionalModel(),
Chris@152:     m_server(model.m_server),
Chris@152:     m_xshift(model.m_xshift),
Chris@152:     m_yshift(model.m_yshift)
Chris@152: {
Chris@152:     FFTDataServer::claimInstance(m_server);
Chris@152: }
Chris@152: