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@35
|
40 m_lowest = 0;
|
Chris@35
|
41 m_highest = m_notes - 1;
|
Chris@35
|
42
|
Chris@35
|
43 for (int i = 0; i < m_instruments; ++i) {
|
Chris@35
|
44 if (i == 0 || silvet_templates[i].lowest < m_lowest) {
|
Chris@35
|
45 m_lowest = silvet_templates[i].lowest;
|
Chris@35
|
46 }
|
Chris@35
|
47 if (i == 0 || silvet_templates[i].highest > m_highest) {
|
Chris@35
|
48 m_highest = silvet_templates[i].highest;
|
Chris@35
|
49 }
|
Chris@35
|
50 }
|
Chris@35
|
51
|
Chris@35
|
52 m_pitches = V(m_notes);
|
Chris@35
|
53
|
Chris@36
|
54 for (int n = 0; n < m_notes; ++n) {
|
Chris@35
|
55 m_pitches[n] = drand48();
|
Chris@35
|
56 }
|
Chris@35
|
57
|
Chris@35
|
58 m_sources = Grid(m_instruments);
|
Chris@35
|
59
|
Chris@35
|
60 for (int i = 0; i < m_instruments; ++i) {
|
Chris@35
|
61 m_sources[i] = V(m_notes);
|
Chris@35
|
62 for (int n = 0; n < m_notes; ++n) {
|
Chris@35
|
63 m_sources[i][n] = (inRange(i, n) ? 1.0 : 0.0);
|
Chris@35
|
64 }
|
Chris@35
|
65 }
|
Chris@35
|
66
|
Chris@36
|
67 m_estimate = V(m_bins);
|
Chris@35
|
68 m_q = V(m_bins);
|
Chris@35
|
69 }
|
Chris@35
|
70
|
Chris@35
|
71 EM::~EM()
|
Chris@35
|
72 {
|
Chris@35
|
73 }
|
Chris@35
|
74
|
Chris@35
|
75 bool
|
Chris@35
|
76 EM::inRange(int instrument, int note)
|
Chris@35
|
77 {
|
Chris@35
|
78 return (note >= silvet_templates[instrument].lowest &&
|
Chris@35
|
79 note <= silvet_templates[instrument].highest);
|
Chris@35
|
80 }
|
Chris@35
|
81
|
Chris@36
|
82 void
|
Chris@36
|
83 EM::normalise(V &column)
|
Chris@36
|
84 {
|
Chris@36
|
85 double sum = 0.0;
|
Chris@36
|
86 for (int i = 0; i < (int)column.size(); ++i) {
|
Chris@36
|
87 sum += column[i];
|
Chris@36
|
88 }
|
Chris@36
|
89 for (int i = 0; i < (int)column.size(); ++i) {
|
Chris@36
|
90 column[i] /= sum;
|
Chris@36
|
91 }
|
Chris@36
|
92 }
|
Chris@36
|
93
|
Chris@36
|
94 void
|
Chris@36
|
95 EM::iterate(V column)
|
Chris@36
|
96 {
|
Chris@36
|
97 normalise(column);
|
Chris@36
|
98 expectation(column);
|
Chris@36
|
99 maximisation(column);
|
Chris@36
|
100 }
|
Chris@36
|
101
|
Chris@36
|
102 void
|
Chris@36
|
103 EM::expectation(const V &column)
|
Chris@36
|
104 {
|
Chris@36
|
105 cerr << ".";
|
Chris@36
|
106
|
Chris@36
|
107 for (int i = 0; i < m_bins; ++i) {
|
Chris@36
|
108 m_estimate[i] = epsilon;
|
Chris@36
|
109 }
|
Chris@36
|
110
|
Chris@36
|
111 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
112 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
113 float *w = silvet_templates[i].data[n];
|
Chris@36
|
114 double pitch = m_pitches[n];
|
Chris@36
|
115 double source = m_sources[i][n];
|
Chris@36
|
116 for (int j = 0; j < m_bins; ++j) {
|
Chris@36
|
117 m_estimate[j] += w[j] * pitch * source;
|
Chris@36
|
118 }
|
Chris@36
|
119 }
|
Chris@36
|
120 }
|
Chris@36
|
121
|
Chris@36
|
122 for (int i = 0; i < m_bins; ++i) {
|
Chris@36
|
123 m_q[i] = column[i] / m_estimate[i];
|
Chris@36
|
124 }
|
Chris@36
|
125 }
|
Chris@36
|
126
|
Chris@36
|
127 void
|
Chris@36
|
128 EM::maximisation(const V &column)
|
Chris@36
|
129 {
|
Chris@36
|
130 V newPitches = m_pitches;
|
Chris@36
|
131
|
Chris@36
|
132 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
133 newPitches[n] = epsilon;
|
Chris@36
|
134 if (n >= m_lowest && n <= m_highest) {
|
Chris@36
|
135 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
136 float *w = silvet_templates[i].data[n];
|
Chris@36
|
137 double pitch = m_pitches[n];
|
Chris@36
|
138 double source = m_sources[i][n];
|
Chris@36
|
139 for (int j = 0; j < m_bins; ++j) {
|
Chris@36
|
140 newPitches[n] += w[j] * m_q[j] * pitch * source;
|
Chris@36
|
141 }
|
Chris@36
|
142 }
|
Chris@36
|
143 }
|
Chris@42
|
144 if (m_pitchSparsity != 1.0) {
|
Chris@42
|
145 newPitches[n] = pow(newPitches[n], m_pitchSparsity);
|
Chris@42
|
146 }
|
Chris@36
|
147 }
|
Chris@36
|
148 normalise(newPitches);
|
Chris@36
|
149
|
Chris@36
|
150 Grid newSources = m_sources;
|
Chris@36
|
151
|
Chris@36
|
152 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
153 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
154 newSources[i][n] = epsilon;
|
Chris@36
|
155 if (inRange(i, n)) {
|
Chris@36
|
156 float *w = silvet_templates[i].data[n];
|
Chris@38
|
157 double pitch = m_pitches[n];
|
Chris@38
|
158 double source = m_sources[i][n];
|
Chris@36
|
159 for (int j = 0; j < m_bins; ++j) {
|
Chris@38
|
160 newSources[i][n] += w[j] * m_q[j] * pitch * source;
|
Chris@36
|
161 }
|
Chris@36
|
162 }
|
Chris@42
|
163 if (m_sourceSparsity != 1.0) {
|
Chris@42
|
164 newSources[i][n] = pow(newSources[i][n], m_sourceSparsity);
|
Chris@42
|
165 }
|
Chris@36
|
166 }
|
Chris@36
|
167 normalise(newSources[i]);
|
Chris@36
|
168 }
|
Chris@36
|
169
|
Chris@36
|
170 m_pitches = newPitches;
|
Chris@36
|
171 m_sources = newSources;
|
Chris@36
|
172 }
|
Chris@36
|
173
|
Chris@36
|
174 void
|
Chris@36
|
175 EM::report()
|
Chris@36
|
176 {
|
Chris@36
|
177 vector<int> sounding;
|
Chris@36
|
178 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
179 if (m_pitches[n] > 0.05) {
|
Chris@36
|
180 sounding.push_back(n);
|
Chris@36
|
181 }
|
Chris@36
|
182 }
|
Chris@36
|
183 cerr << " sounding: ";
|
Chris@36
|
184 for (int i = 0; i < (int)sounding.size(); ++i) {
|
Chris@36
|
185 cerr << sounding[i] << " ";
|
Chris@37
|
186 int maxj = -1;
|
Chris@37
|
187 double maxs = 0.0;
|
Chris@37
|
188 for (int j = 0; j < m_instruments; ++j) {
|
Chris@37
|
189 if (j == 0 || m_sources[j][sounding[i]] > maxs) {
|
Chris@37
|
190 maxj = j;
|
Chris@37
|
191 maxs = m_sources[j][sounding[i]];
|
Chris@37
|
192 }
|
Chris@37
|
193 }
|
Chris@37
|
194 cerr << silvet_templates[maxj].name << " ";
|
Chris@36
|
195 }
|
Chris@36
|
196 cerr << endl;
|
Chris@36
|
197 }
|
Chris@36
|
198
|