Chris@34: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
Chris@34: 
Chris@34: /*
Chris@34:   Silvet
Chris@34: 
Chris@34:   A Vamp plugin for note transcription.
Chris@34:   Centre for Digital Music, Queen Mary University of London.
Chris@34:     
Chris@34:   This program is free software; you can redistribute it and/or
Chris@34:   modify it under the terms of the GNU General Public License as
Chris@34:   published by the Free Software Foundation; either version 2 of the
Chris@34:   License, or (at your option) any later version.  See the file
Chris@34:   COPYING included with this distribution for more information.
Chris@34: */
Chris@34: 
Chris@34: #include "EM.h"
Chris@34: 
Chris@34: #include "data/include/templates.h"
Chris@34: 
Chris@36: #include <cstdlib>
Chris@42: #include <cmath>
Chris@36: 
Chris@36: #include <iostream>
Chris@36: 
Chris@36: #include <vector>
Chris@36: 
Chris@36: using std::vector;
Chris@36: using std::cerr;
Chris@36: using std::endl;
Chris@36: 
Chris@35: static double epsilon = 1e-16;
Chris@35: 
Chris@35: EM::EM() :
Chris@45:     m_noteCount(SILVET_TEMPLATE_NOTE_COUNT),
Chris@45:     m_shiftCount(SILVET_TEMPLATE_MAX_SHIFT * 2 + 1),
Chris@45:     m_binCount(SILVET_TEMPLATE_HEIGHT),
Chris@45:     m_instrumentCount(SILVET_TEMPLATE_COUNT),
Chris@42:     m_pitchSparsity(1.1),
Chris@83:     m_sourceSparsity(1.3),
Chris@83:     m_lowestPitch(silvet_templates_lowest_note),
Chris@83:     m_highestPitch(silvet_templates_highest_note)
Chris@35: {
Chris@55:     m_pitches = V(m_noteCount);
Chris@55:     for (int n = 0; n < m_noteCount; ++n) {
Chris@55:         m_pitches[n] = drand48();
Chris@55:     }
Chris@35: 
Chris@55:     m_shifts = Grid(m_shiftCount);
Chris@55:     for (int f = 0; f < m_shiftCount; ++f) {
Chris@55:         m_shifts[f] = V(m_noteCount);
Chris@55:         for (int n = 0; n < m_noteCount; ++n) {
Chris@55:             m_shifts[f][n] = drand48();
Chris@55:         }
Chris@35:     }
Chris@35:     
Chris@45:     m_sources = Grid(m_instrumentCount);
Chris@55:         for (int i = 0; i < m_instrumentCount; ++i) {
Chris@55:         m_sources[i] = V(m_noteCount);
Chris@55:         for (int n = 0; n < m_noteCount; ++n) {
Chris@35:             m_sources[i][n] = (inRange(i, n) ? 1.0 : 0.0);
Chris@35:         }
Chris@35:     }
Chris@35: 
Chris@45:     m_estimate = V(m_binCount);
Chris@45:     m_q = V(m_binCount);
Chris@35: }
Chris@35: 
Chris@35: EM::~EM()
Chris@35: {
Chris@35: }
Chris@35: 
Chris@45: void
Chris@45: EM::rangeFor(int instrument, int &minPitch, int &maxPitch)
Chris@45: {
Chris@55:     minPitch = silvet_templates[instrument].lowest;
Chris@55:     maxPitch = silvet_templates[instrument].highest;
Chris@45: }
Chris@45: 
Chris@35: bool
Chris@45: EM::inRange(int instrument, int pitch)
Chris@35: {
Chris@45:     int minPitch, maxPitch;
Chris@45:     rangeFor(instrument, minPitch, maxPitch);
Chris@45:     return (pitch >= minPitch && pitch <= maxPitch);
Chris@35: }
Chris@35: 
Chris@36: void
Chris@55: EM::normaliseColumn(V &column)
Chris@36: {
Chris@36:     double sum = 0.0;
Chris@36:     for (int i = 0; i < (int)column.size(); ++i) {
Chris@36:         sum += column[i];
Chris@36:     }
Chris@36:     for (int i = 0; i < (int)column.size(); ++i) {
Chris@36:         column[i] /= sum;
Chris@36:     }
Chris@36: }
Chris@36: 
Chris@36: void
Chris@55: EM::normaliseGrid(Grid &grid)
Chris@53: {
Chris@55:     V denominators(grid[0].size());
Chris@53: 
Chris@55:     for (int i = 0; i < (int)grid.size(); ++i) {
Chris@55:         for (int j = 0; j < (int)grid[i].size(); ++j) {
Chris@55:             denominators[j] += grid[i][j];
Chris@53:         }
Chris@53:     }
Chris@53: 
Chris@55:     for (int i = 0; i < (int)grid.size(); ++i) {
Chris@55:         for (int j = 0; j < (int)grid[i].size(); ++j) {
Chris@55:             grid[i][j] /= denominators[j];
Chris@53:         }
Chris@53:     }
Chris@53: }
Chris@53: 
Chris@53: void
Chris@36: EM::iterate(V column)
Chris@36: {
Chris@55:     normaliseColumn(column);
Chris@36:     expectation(column);
Chris@36:     maximisation(column);
Chris@36: }
Chris@36: 
Chris@45: const float *
Chris@55: EM::templateFor(int instrument, int note, int shift)
Chris@45: {
Chris@45:     return silvet_templates[instrument].data[note] + shift;
Chris@45: }
Chris@45: 
Chris@36: void
Chris@36: EM::expectation(const V &column)
Chris@36: {
Chris@62: //    cerr << ".";
Chris@36: 
Chris@45:     for (int i = 0; i < m_binCount; ++i) {
Chris@36:         m_estimate[i] = epsilon;
Chris@36:     }
Chris@36: 
Chris@45:     for (int i = 0; i < m_instrumentCount; ++i) {
Chris@55:         for (int n = 0; n < m_noteCount; ++n) {
Chris@83:             const double pitch = m_pitches[n];
Chris@83:             const double source = m_sources[i][n];
Chris@55:             for (int f = 0; f < m_shiftCount; ++f) {
Chris@55:                 const float *w = templateFor(i, n, f);
Chris@83:                 const double shift = m_shifts[f][n];
Chris@83:                 const double factor = pitch * source * shift;
Chris@55:                 for (int j = 0; j < m_binCount; ++j) {
Chris@83:                     m_estimate[j] += w[j] * factor;
Chris@55:                 }
Chris@36:             }
Chris@36:         }
Chris@36:     }
Chris@36: 
Chris@45:     for (int i = 0; i < m_binCount; ++i) {
Chris@36:         m_q[i] = column[i] / m_estimate[i];
Chris@36:     }
Chris@36: }
Chris@36: 
Chris@36: void
Chris@36: EM::maximisation(const V &column)
Chris@36: {
Chris@36:     V newPitches = m_pitches;
Chris@36: 
Chris@55:     for (int n = 0; n < m_noteCount; ++n) {
Chris@36:         newPitches[n] = epsilon;
Chris@45:         if (n >= m_lowestPitch && n <= m_highestPitch) {
Chris@83:             const double pitch = m_pitches[n];
Chris@45:             for (int i = 0; i < m_instrumentCount; ++i) {
Chris@83:                 const double source = m_sources[i][n];
Chris@55:                 for (int f = 0; f < m_shiftCount; ++f) {
Chris@55:                     const float *w = templateFor(i, n, f);
Chris@83:                     const double shift = m_shifts[f][n];
Chris@83:                     const double factor = pitch * source * shift;
Chris@55:                     for (int j = 0; j < m_binCount; ++j) {
Chris@83:                         newPitches[n] += w[j] * m_q[j] * factor;
Chris@55:                     }
Chris@36:                 }
Chris@36:             }
Chris@36:         }
Chris@42:         if (m_pitchSparsity != 1.0) {
Chris@42:             newPitches[n] = pow(newPitches[n], m_pitchSparsity);
Chris@42:         }
Chris@36:     }
Chris@55:     normaliseColumn(newPitches);
Chris@55: 
Chris@55:     Grid newShifts = m_shifts;
Chris@55: 
Chris@55:     for (int f = 0; f < m_shiftCount; ++f) {
Chris@55:         for (int n = 0; n < m_noteCount; ++n) {
Chris@83:             const double pitch = m_pitches[n];
Chris@83:             const double shift = m_shifts[f][n];
Chris@55:             newShifts[f][n] = epsilon;
Chris@55:             for (int i = 0; i < m_instrumentCount; ++i) {
Chris@55:                 const float *w = templateFor(i, n, f);
Chris@83:                 const double source = m_sources[i][n];
Chris@83:                 const double factor = pitch * source * shift;
Chris@55:                 for (int j = 0; j < m_binCount; ++j) {
Chris@83:                     newShifts[f][n] += w[j] * m_q[j] * factor;
Chris@55:                 }
Chris@55:             }
Chris@55:         }
Chris@55:     }
Chris@55:     normaliseGrid(newShifts);
Chris@36: 
Chris@36:     Grid newSources = m_sources;
Chris@36: 
Chris@45:     for (int i = 0; i < m_instrumentCount; ++i) {
Chris@55:         for (int n = 0; n < m_noteCount; ++n) {
Chris@83:             const double pitch = m_pitches[n];
Chris@83:             const double source = m_sources[i][n];
Chris@36:             newSources[i][n] = epsilon;
Chris@36:             if (inRange(i, n)) {
Chris@55:                 for (int f = 0; f < m_shiftCount; ++f) {
Chris@55:                     const float *w = templateFor(i, n, f);
Chris@83:                     const double shift = m_shifts[f][n];
Chris@83:                     const double factor = pitch * source * shift;
Chris@55:                     for (int j = 0; j < m_binCount; ++j) {
Chris@83:                         newSources[i][n] += w[j] * m_q[j] * factor;
Chris@55:                     }
Chris@36:                 }
Chris@36:             }
Chris@42:             if (m_sourceSparsity != 1.0) {
Chris@42:                 newSources[i][n] = pow(newSources[i][n], m_sourceSparsity);
Chris@42:             }
Chris@36:         }
Chris@36:     }
Chris@55:     normaliseGrid(newSources);
Chris@36: 
Chris@36:     m_pitches = newPitches;
Chris@55:     m_shifts = newShifts;
Chris@36:     m_sources = newSources;
Chris@36: }
Chris@36: 
Chris@36: