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@36
|
21
|
Chris@36
|
22 #include <iostream>
|
Chris@36
|
23
|
Chris@36
|
24 #include <vector>
|
Chris@36
|
25
|
Chris@36
|
26 using std::vector;
|
Chris@36
|
27 using std::cerr;
|
Chris@36
|
28 using std::endl;
|
Chris@36
|
29
|
Chris@35
|
30 static double epsilon = 1e-16;
|
Chris@35
|
31
|
Chris@35
|
32 EM::EM() :
|
Chris@35
|
33 m_notes(SILVET_TEMPLATE_NOTE_COUNT),
|
Chris@35
|
34 m_bins(SILVET_TEMPLATE_HEIGHT),
|
Chris@35
|
35 m_instruments(SILVET_TEMPLATE_COUNT)
|
Chris@35
|
36 {
|
Chris@36
|
37 cerr << "init!" << endl;
|
Chris@36
|
38
|
Chris@35
|
39 m_lowest = 0;
|
Chris@35
|
40 m_highest = m_notes - 1;
|
Chris@35
|
41
|
Chris@35
|
42 for (int i = 0; i < m_instruments; ++i) {
|
Chris@35
|
43 if (i == 0 || silvet_templates[i].lowest < m_lowest) {
|
Chris@35
|
44 m_lowest = silvet_templates[i].lowest;
|
Chris@35
|
45 }
|
Chris@35
|
46 if (i == 0 || silvet_templates[i].highest > m_highest) {
|
Chris@35
|
47 m_highest = silvet_templates[i].highest;
|
Chris@35
|
48 }
|
Chris@35
|
49 }
|
Chris@35
|
50
|
Chris@35
|
51 m_pitches = V(m_notes);
|
Chris@35
|
52
|
Chris@36
|
53 for (int n = 0; n < m_notes; ++n) {
|
Chris@35
|
54 m_pitches[n] = drand48();
|
Chris@35
|
55 }
|
Chris@35
|
56
|
Chris@35
|
57 m_sources = Grid(m_instruments);
|
Chris@35
|
58
|
Chris@35
|
59 for (int i = 0; i < m_instruments; ++i) {
|
Chris@35
|
60 m_sources[i] = V(m_notes);
|
Chris@35
|
61 for (int n = 0; n < m_notes; ++n) {
|
Chris@35
|
62 m_sources[i][n] = (inRange(i, n) ? 1.0 : 0.0);
|
Chris@35
|
63 }
|
Chris@35
|
64 }
|
Chris@35
|
65
|
Chris@36
|
66 m_estimate = V(m_bins);
|
Chris@35
|
67 m_q = V(m_bins);
|
Chris@36
|
68
|
Chris@36
|
69 cerr << "(init done)" << endl;
|
Chris@35
|
70 }
|
Chris@35
|
71
|
Chris@35
|
72 EM::~EM()
|
Chris@35
|
73 {
|
Chris@35
|
74 }
|
Chris@35
|
75
|
Chris@35
|
76 bool
|
Chris@35
|
77 EM::inRange(int instrument, int note)
|
Chris@35
|
78 {
|
Chris@35
|
79 return (note >= silvet_templates[instrument].lowest &&
|
Chris@35
|
80 note <= silvet_templates[instrument].highest);
|
Chris@35
|
81 }
|
Chris@35
|
82
|
Chris@36
|
83 void
|
Chris@36
|
84 EM::normalise(V &column)
|
Chris@36
|
85 {
|
Chris@36
|
86 double sum = 0.0;
|
Chris@36
|
87 for (int i = 0; i < (int)column.size(); ++i) {
|
Chris@36
|
88 sum += column[i];
|
Chris@36
|
89 }
|
Chris@36
|
90 for (int i = 0; i < (int)column.size(); ++i) {
|
Chris@36
|
91 column[i] /= sum;
|
Chris@36
|
92 }
|
Chris@36
|
93 }
|
Chris@36
|
94
|
Chris@36
|
95 void
|
Chris@36
|
96 EM::iterate(V column)
|
Chris@36
|
97 {
|
Chris@36
|
98 normalise(column);
|
Chris@36
|
99 expectation(column);
|
Chris@36
|
100 maximisation(column);
|
Chris@36
|
101 }
|
Chris@36
|
102
|
Chris@36
|
103 void
|
Chris@36
|
104 EM::expectation(const V &column)
|
Chris@36
|
105 {
|
Chris@36
|
106 cerr << ".";
|
Chris@36
|
107
|
Chris@36
|
108 for (int i = 0; i < m_bins; ++i) {
|
Chris@36
|
109 m_estimate[i] = epsilon;
|
Chris@36
|
110 }
|
Chris@36
|
111
|
Chris@36
|
112 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
113 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
114 float *w = silvet_templates[i].data[n];
|
Chris@36
|
115 double pitch = m_pitches[n];
|
Chris@36
|
116 double source = m_sources[i][n];
|
Chris@36
|
117 for (int j = 0; j < m_bins; ++j) {
|
Chris@36
|
118 m_estimate[j] += w[j] * pitch * source;
|
Chris@36
|
119 }
|
Chris@36
|
120 }
|
Chris@36
|
121 }
|
Chris@36
|
122
|
Chris@36
|
123 for (int i = 0; i < m_bins; ++i) {
|
Chris@36
|
124 m_q[i] = column[i] / m_estimate[i];
|
Chris@36
|
125 }
|
Chris@36
|
126 }
|
Chris@36
|
127
|
Chris@36
|
128 void
|
Chris@36
|
129 EM::maximisation(const V &column)
|
Chris@36
|
130 {
|
Chris@36
|
131 V newPitches = m_pitches;
|
Chris@36
|
132
|
Chris@36
|
133 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
134 newPitches[n] = epsilon;
|
Chris@36
|
135 if (n >= m_lowest && n <= m_highest) {
|
Chris@36
|
136 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
137 float *w = silvet_templates[i].data[n];
|
Chris@36
|
138 double pitch = m_pitches[n];
|
Chris@36
|
139 double source = m_sources[i][n];
|
Chris@36
|
140 for (int j = 0; j < m_bins; ++j) {
|
Chris@36
|
141 newPitches[n] += w[j] * m_q[j] * pitch * source;
|
Chris@36
|
142 }
|
Chris@36
|
143 }
|
Chris@36
|
144 }
|
Chris@36
|
145 }
|
Chris@36
|
146 normalise(newPitches);
|
Chris@36
|
147
|
Chris@36
|
148 Grid newSources = m_sources;
|
Chris@36
|
149
|
Chris@36
|
150 for (int i = 0; i < m_instruments; ++i) {
|
Chris@36
|
151 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
152 newSources[i][n] = epsilon;
|
Chris@36
|
153 if (inRange(i, n)) {
|
Chris@36
|
154 float *w = silvet_templates[i].data[n];
|
Chris@36
|
155 for (int j = 0; j < m_bins; ++j) {
|
Chris@36
|
156 newSources[i][n] +=
|
Chris@36
|
157 w[j] * m_q[j] * m_pitches[n] * m_sources[i][n];
|
Chris@36
|
158 }
|
Chris@36
|
159 }
|
Chris@36
|
160 }
|
Chris@36
|
161 normalise(newSources[i]);
|
Chris@36
|
162 }
|
Chris@36
|
163
|
Chris@36
|
164 m_pitches = newPitches;
|
Chris@36
|
165 m_sources = newSources;
|
Chris@36
|
166 }
|
Chris@36
|
167
|
Chris@36
|
168 void
|
Chris@36
|
169 EM::report()
|
Chris@36
|
170 {
|
Chris@36
|
171 vector<int> sounding;
|
Chris@36
|
172 for (int n = 0; n < m_notes; ++n) {
|
Chris@36
|
173 if (m_pitches[n] > 0.05) {
|
Chris@36
|
174 sounding.push_back(n);
|
Chris@36
|
175 }
|
Chris@36
|
176 }
|
Chris@36
|
177 cerr << " sounding: ";
|
Chris@36
|
178 for (int i = 0; i < (int)sounding.size(); ++i) {
|
Chris@36
|
179 cerr << sounding[i] << " ";
|
Chris@37
|
180 int maxj = -1;
|
Chris@37
|
181 double maxs = 0.0;
|
Chris@37
|
182 for (int j = 0; j < m_instruments; ++j) {
|
Chris@37
|
183 if (j == 0 || m_sources[j][sounding[i]] > maxs) {
|
Chris@37
|
184 maxj = j;
|
Chris@37
|
185 maxs = m_sources[j][sounding[i]];
|
Chris@37
|
186 }
|
Chris@37
|
187 }
|
Chris@37
|
188 cerr << silvet_templates[maxj].name << " ";
|
Chris@36
|
189 }
|
Chris@36
|
190 cerr << endl;
|
Chris@36
|
191 }
|
Chris@36
|
192
|