Chris@34
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
Chris@34
|
2
|
Chris@34
|
3 /*
|
Chris@34
|
4 Silvet
|
Chris@34
|
5
|
Chris@34
|
6 A Vamp plugin for note transcription.
|
Chris@34
|
7 Centre for Digital Music, Queen Mary University of London.
|
Chris@34
|
8
|
Chris@34
|
9 This program is free software; you can redistribute it and/or
|
Chris@34
|
10 modify it under the terms of the GNU General Public License as
|
Chris@34
|
11 published by the Free Software Foundation; either version 2 of the
|
Chris@34
|
12 License, or (at your option) any later version. See the file
|
Chris@34
|
13 COPYING included with this distribution for more information.
|
Chris@34
|
14 */
|
Chris@34
|
15
|
Chris@34
|
16 #include "EM.h"
|
Chris@34
|
17
|
Chris@34
|
18 #include "data/include/templates.h"
|
Chris@34
|
19
|
Chris@36
|
20 #include <cstdlib>
|
Chris@42
|
21 #include <cmath>
|
Chris@36
|
22
|
Chris@36
|
23 #include <iostream>
|
Chris@36
|
24
|
Chris@36
|
25 #include <vector>
|
Chris@36
|
26
|
Chris@36
|
27 using std::vector;
|
Chris@36
|
28 using std::cerr;
|
Chris@36
|
29 using std::endl;
|
Chris@36
|
30
|
Chris@35
|
31 static double epsilon = 1e-16;
|
Chris@35
|
32
|
Chris@35
|
33 EM::EM() :
|
Chris@35
|
34 m_notes(SILVET_TEMPLATE_NOTE_COUNT),
|
Chris@35
|
35 m_bins(SILVET_TEMPLATE_HEIGHT),
|
Chris@42
|
36 m_instruments(SILVET_TEMPLATE_COUNT),
|
Chris@42
|
37 m_pitchSparsity(1.1),
|
Chris@42
|
38 m_sourceSparsity(1.3)
|
Chris@35
|
39 {
|
Chris@49
|
40 m_lowest = silvet_templates_lowest_note;
|
Chris@49
|
41 m_highest = silvet_templates_highest_note;
|
Chris@35
|
42
|
Chris@35
|
43 m_pitches = V(m_notes);
|
Chris@35
|
44
|
Chris@36
|
45 for (int n = 0; n < m_notes; ++n) {
|
Chris@35
|
46 m_pitches[n] = drand48();
|
Chris@35
|
47 }
|
Chris@35
|
48
|
Chris@35
|
49 m_sources = Grid(m_instruments);
|
Chris@35
|
50
|
Chris@35
|
51 for (int i = 0; i < m_instruments; ++i) {
|
Chris@35
|
52 m_sources[i] = V(m_notes);
|
Chris@35
|
53 for (int n = 0; n < m_notes; ++n) {
|
Chris@35
|
54 m_sources[i][n] = (inRange(i, n) ? 1.0 : 0.0);
|
Chris@35
|
55 }
|
Chris@35
|
56 }
|
Chris@35
|
57
|
Chris@36
|
58 m_estimate = V(m_bins);
|
Chris@35
|
59 m_q = V(m_bins);
|
Chris@35
|
60 }
|
Chris@35
|
61
|
Chris@35
|
62 EM::~EM()
|
Chris@35
|
63 {
|
Chris@35
|
64 }
|
Chris@35
|
65
|
Chris@35
|
66 bool
|
Chris@35
|
67 EM::inRange(int instrument, int note)
|
Chris@35
|
68 {
|
Chris@35
|
69 return (note >= silvet_templates[instrument].lowest &&
|
Chris@35
|
70 note <= silvet_templates[instrument].highest);
|
Chris@35
|
71 }
|
Chris@35
|
72
|
Chris@36
|
73 void
|
Chris@36
|
74 EM::normalise(V &column)
|
Chris@36
|
75 {
|
Chris@36
|
76 double sum = 0.0;
|
Chris@36
|
77 for (int i = 0; i < (int)column.size(); ++i) {
|
Chris@36
|
78 sum += column[i];
|
Chris@36
|
79 }
|
Chris@36
|
80 for (int i = 0; i < (int)column.size(); ++i) {
|
Chris@36
|
81 column[i] /= sum;
|
Chris@36
|
82 }
|
Chris@36
|
83 }
|
Chris@36
|
84
|
Chris@36
|
85 void
|
Chris@36
|
86 EM::iterate(V column)
|
Chris@36
|
87 {
|
Chris@36
|
88 normalise(column);
|
Chris@36
|
89 expectation(column);
|
Chris@36
|
90 maximisation(column);
|
Chris@36
|
91 }
|
Chris@36
|
92
|
Chris@36
|
93 void
|
Chris@36
|
94 EM::expectation(const V &column)
|
Chris@36
|
95 {
|
Chris@36
|
96 cerr << ".";
|
Chris@36
|
97
|
Chris@36
|
98 for (int i = 0; i < m_bins; ++i) {
|
Chris@36
|
99 m_estimate[i] = epsilon;
|
Chris@36
|
100 }
|
Chris@36
|
101
|
Chris@36
|
102 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
103 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
104 float *w = silvet_templates[i].data[n];
|
Chris@36
|
105 double pitch = m_pitches[n];
|
Chris@36
|
106 double source = m_sources[i][n];
|
Chris@36
|
107 for (int j = 0; j < m_bins; ++j) {
|
Chris@36
|
108 m_estimate[j] += w[j] * pitch * source;
|
Chris@36
|
109 }
|
Chris@36
|
110 }
|
Chris@36
|
111 }
|
Chris@36
|
112
|
Chris@36
|
113 for (int i = 0; i < m_bins; ++i) {
|
Chris@36
|
114 m_q[i] = column[i] / m_estimate[i];
|
Chris@36
|
115 }
|
Chris@36
|
116 }
|
Chris@36
|
117
|
Chris@36
|
118 void
|
Chris@36
|
119 EM::maximisation(const V &column)
|
Chris@36
|
120 {
|
Chris@36
|
121 V newPitches = m_pitches;
|
Chris@36
|
122
|
Chris@36
|
123 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
124 newPitches[n] = epsilon;
|
Chris@36
|
125 if (n >= m_lowest && n <= m_highest) {
|
Chris@36
|
126 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
127 float *w = silvet_templates[i].data[n];
|
Chris@36
|
128 double pitch = m_pitches[n];
|
Chris@36
|
129 double source = m_sources[i][n];
|
Chris@36
|
130 for (int j = 0; j < m_bins; ++j) {
|
Chris@36
|
131 newPitches[n] += w[j] * m_q[j] * pitch * source;
|
Chris@36
|
132 }
|
Chris@36
|
133 }
|
Chris@36
|
134 }
|
Chris@42
|
135 if (m_pitchSparsity != 1.0) {
|
Chris@42
|
136 newPitches[n] = pow(newPitches[n], m_pitchSparsity);
|
Chris@42
|
137 }
|
Chris@36
|
138 }
|
Chris@36
|
139 normalise(newPitches);
|
Chris@36
|
140
|
Chris@36
|
141 Grid newSources = m_sources;
|
Chris@36
|
142
|
Chris@36
|
143 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
144 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
145 newSources[i][n] = epsilon;
|
Chris@36
|
146 if (inRange(i, n)) {
|
Chris@36
|
147 float *w = silvet_templates[i].data[n];
|
Chris@38
|
148 double pitch = m_pitches[n];
|
Chris@38
|
149 double source = m_sources[i][n];
|
Chris@36
|
150 for (int j = 0; j < m_bins; ++j) {
|
Chris@38
|
151 newSources[i][n] += w[j] * m_q[j] * pitch * source;
|
Chris@36
|
152 }
|
Chris@36
|
153 }
|
Chris@42
|
154 if (m_sourceSparsity != 1.0) {
|
Chris@42
|
155 newSources[i][n] = pow(newSources[i][n], m_sourceSparsity);
|
Chris@42
|
156 }
|
Chris@36
|
157 }
|
Chris@36
|
158 normalise(newSources[i]);
|
Chris@36
|
159 }
|
Chris@36
|
160
|
Chris@36
|
161 m_pitches = newPitches;
|
Chris@36
|
162 m_sources = newSources;
|
Chris@36
|
163 }
|
Chris@36
|
164
|
Chris@36
|
165 void
|
Chris@36
|
166 EM::report()
|
Chris@36
|
167 {
|
Chris@36
|
168 vector<int> sounding;
|
Chris@36
|
169 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
170 if (m_pitches[n] > 0.05) {
|
Chris@36
|
171 sounding.push_back(n);
|
Chris@36
|
172 }
|
Chris@36
|
173 }
|
Chris@36
|
174 cerr << " sounding: ";
|
Chris@36
|
175 for (int i = 0; i < (int)sounding.size(); ++i) {
|
Chris@36
|
176 cerr << sounding[i] << " ";
|
Chris@37
|
177 int maxj = -1;
|
Chris@37
|
178 double maxs = 0.0;
|
Chris@37
|
179 for (int j = 0; j < m_instruments; ++j) {
|
Chris@37
|
180 if (j == 0 || m_sources[j][sounding[i]] > maxs) {
|
Chris@37
|
181 maxj = j;
|
Chris@37
|
182 maxs = m_sources[j][sounding[i]];
|
Chris@37
|
183 }
|
Chris@37
|
184 }
|
Chris@37
|
185 cerr << silvet_templates[maxj].name << " ";
|
Chris@36
|
186 }
|
Chris@36
|
187 cerr << endl;
|
Chris@36
|
188 }
|
Chris@36
|
189
|