cannam@475
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
cannam@475
|
2 /*
|
cannam@475
|
3 QM DSP Library
|
cannam@475
|
4
|
cannam@475
|
5 Centre for Digital Music, Queen Mary, University of London.
|
cannam@475
|
6 This file 2005-2006 Christian Landone and Katy Noland.
|
cannam@475
|
7
|
cannam@475
|
8 Fixes to correct chroma offsets and for thread safety contributed
|
cannam@475
|
9 by Daniel Schürmann.
|
cannam@475
|
10
|
cannam@475
|
11 This program is free software; you can redistribute it and/or
|
cannam@475
|
12 modify it under the terms of the GNU General Public License as
|
cannam@475
|
13 published by the Free Software Foundation; either version 2 of the
|
cannam@475
|
14 License, or (at your option) any later version. See the file
|
cannam@475
|
15 COPYING included with this distribution for more information.
|
cannam@475
|
16 */
|
cannam@475
|
17
|
cannam@475
|
18 #include "GetKeyMode.h"
|
cannam@475
|
19 #include "maths/MathUtilities.h"
|
cannam@475
|
20 #include "base/Pitch.h"
|
cannam@475
|
21
|
cannam@475
|
22 #include <iostream>
|
cannam@475
|
23
|
cannam@475
|
24 #include <cstring>
|
cannam@475
|
25 #include <cstdlib>
|
cannam@475
|
26
|
cannam@475
|
27 static const int kBinsPerOctave = 36;
|
cannam@475
|
28
|
cannam@475
|
29 // Chords profile
|
cannam@475
|
30 static double MajProfile[kBinsPerOctave] = {
|
cannam@475
|
31 0.0384, 0.0629, 0.0258, 0.0121, 0.0146, 0.0106, 0.0364, 0.0610, 0.0267,
|
cannam@475
|
32 0.0126, 0.0121, 0.0086, 0.0364, 0.0623, 0.0279, 0.0275, 0.0414, 0.0186,
|
cannam@475
|
33 0.0173, 0.0248, 0.0145, 0.0364, 0.0631, 0.0262, 0.0129, 0.0150, 0.0098,
|
cannam@475
|
34 0.0312, 0.0521, 0.0235, 0.0129, 0.0142, 0.0095, 0.0289, 0.0478, 0.0239};
|
cannam@475
|
35
|
cannam@475
|
36 static double MinProfile[kBinsPerOctave] = {
|
cannam@475
|
37 0.0375, 0.0682, 0.0299, 0.0119, 0.0138, 0.0093, 0.0296, 0.0543, 0.0257,
|
cannam@475
|
38 0.0292, 0.0519, 0.0246, 0.0159, 0.0234, 0.0135, 0.0291, 0.0544, 0.0248,
|
cannam@475
|
39 0.0137, 0.0176, 0.0104, 0.0352, 0.0670, 0.0302, 0.0222, 0.0349, 0.0164,
|
cannam@475
|
40 0.0174, 0.0297, 0.0166, 0.0222, 0.0401, 0.0202, 0.0175, 0.0270, 0.0146};
|
cannam@475
|
41 //
|
cannam@475
|
42
|
cannam@475
|
43
|
cannam@475
|
44 //////////////////////////////////////////////////////////////////////
|
cannam@475
|
45 // Construction/Destruction
|
cannam@475
|
46 //////////////////////////////////////////////////////////////////////
|
cannam@475
|
47
|
cannam@475
|
48 GetKeyMode::GetKeyMode( int sampleRate, float tuningFrequency,
|
cannam@475
|
49 double hpcpAverage, double medianAverage ) :
|
cannam@475
|
50 m_hpcpAverage( hpcpAverage ),
|
cannam@475
|
51 m_medianAverage( medianAverage ),
|
cannam@475
|
52 m_ChrPointer(0),
|
cannam@475
|
53 m_DecimatedBuffer(0),
|
cannam@475
|
54 m_ChromaBuffer(0),
|
cannam@475
|
55 m_MeanHPCP(0),
|
cannam@475
|
56 m_MajCorr(0),
|
cannam@475
|
57 m_MinCorr(0),
|
cannam@475
|
58 m_MedianFilterBuffer(0),
|
cannam@475
|
59 m_SortedBuffer(0),
|
cannam@475
|
60 m_keyStrengths(0)
|
cannam@475
|
61 {
|
cannam@475
|
62 m_DecimationFactor = 8;
|
cannam@475
|
63
|
cannam@475
|
64 // Chromagram configuration parameters
|
cannam@475
|
65 m_ChromaConfig.normalise = MathUtilities::NormaliseUnitMax;
|
cannam@475
|
66 m_ChromaConfig.FS = sampleRate/(double)m_DecimationFactor;
|
cannam@475
|
67 if (m_ChromaConfig.FS < 1) {
|
cannam@475
|
68 m_ChromaConfig.FS = 1;
|
cannam@475
|
69 }
|
cannam@475
|
70
|
cannam@475
|
71 // Set C3 (= MIDI #48) as our base:
|
cannam@475
|
72 // This implies that key = 1 => Cmaj, key = 12 => Bmaj, key = 13 => Cmin, etc.
|
cannam@475
|
73 m_ChromaConfig.min = Pitch::getFrequencyForPitch( 48, 0, tuningFrequency );
|
cannam@475
|
74 m_ChromaConfig.max = Pitch::getFrequencyForPitch( 96, 0, tuningFrequency );
|
cannam@475
|
75
|
cannam@475
|
76 m_ChromaConfig.BPO = kBinsPerOctave;
|
cannam@475
|
77 m_ChromaConfig.CQThresh = 0.0054;
|
cannam@475
|
78
|
cannam@475
|
79 // Chromagram inst.
|
cannam@475
|
80 m_Chroma = new Chromagram( m_ChromaConfig );
|
cannam@475
|
81
|
cannam@475
|
82 // Get calculated parameters from chroma object
|
cannam@475
|
83 m_ChromaFrameSize = m_Chroma->getFrameSize();
|
cannam@475
|
84 // override hopsize for this application
|
cannam@475
|
85 m_ChromaHopSize = m_ChromaFrameSize;
|
cannam@475
|
86
|
cannam@475
|
87 // std::cerr << "chroma frame size = " << m_ChromaFrameSize << ", decimation factor = " << m_DecimationFactor << " therefore block size = " << getBlockSize() << std::endl;
|
cannam@475
|
88
|
cannam@475
|
89 // Chromagram average and estimated key median filter lengths
|
cannam@475
|
90 m_ChromaBuffersize = (int)ceil( m_hpcpAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
|
cannam@475
|
91 m_MedianWinsize = (int)ceil( m_medianAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
|
cannam@475
|
92
|
cannam@475
|
93 // Reset counters
|
cannam@475
|
94 m_bufferindex = 0;
|
cannam@475
|
95 m_ChromaBufferFilling = 0;
|
cannam@475
|
96 m_MedianBufferFilling = 0;
|
cannam@475
|
97
|
cannam@475
|
98 // Spawn objectc/arrays
|
cannam@475
|
99 m_DecimatedBuffer = new double[m_ChromaFrameSize];
|
cannam@475
|
100
|
cannam@475
|
101 m_ChromaBuffer = new double[kBinsPerOctave * m_ChromaBuffersize];
|
cannam@475
|
102 memset( m_ChromaBuffer, 0, sizeof(double) * kBinsPerOctave * m_ChromaBuffersize);
|
cannam@475
|
103
|
cannam@475
|
104 m_MeanHPCP = new double[kBinsPerOctave];
|
cannam@475
|
105
|
cannam@475
|
106 m_MajCorr = new double[kBinsPerOctave];
|
cannam@475
|
107 m_MinCorr = new double[kBinsPerOctave];
|
cannam@475
|
108
|
cannam@475
|
109 m_MajProfileNorm = new double[kBinsPerOctave];
|
cannam@475
|
110 m_MinProfileNorm = new double[kBinsPerOctave];
|
cannam@475
|
111
|
cannam@475
|
112 double mMaj = MathUtilities::mean( MajProfile, kBinsPerOctave );
|
cannam@475
|
113 double mMin = MathUtilities::mean( MinProfile, kBinsPerOctave );
|
cannam@475
|
114
|
cannam@475
|
115 for( unsigned int i = 0; i < kBinsPerOctave; i++ ) {
|
cannam@475
|
116 m_MajProfileNorm[i] = MajProfile[i] - mMaj;
|
cannam@475
|
117 m_MinProfileNorm[i] = MinProfile[i] - mMin;
|
cannam@475
|
118 }
|
cannam@475
|
119
|
cannam@475
|
120 m_MedianFilterBuffer = new int[ m_MedianWinsize ];
|
cannam@475
|
121 memset( m_MedianFilterBuffer, 0, sizeof(int)*m_MedianWinsize);
|
cannam@475
|
122
|
cannam@475
|
123 m_SortedBuffer = new int[ m_MedianWinsize ];
|
cannam@475
|
124 memset( m_SortedBuffer, 0, sizeof(int)*m_MedianWinsize);
|
cannam@475
|
125
|
cannam@475
|
126 m_Decimator = new Decimator( m_ChromaFrameSize*m_DecimationFactor, m_DecimationFactor );
|
cannam@475
|
127
|
cannam@475
|
128 m_keyStrengths = new double[24];
|
cannam@475
|
129 }
|
cannam@475
|
130
|
cannam@475
|
131 GetKeyMode::~GetKeyMode()
|
cannam@475
|
132 {
|
cannam@475
|
133 delete m_Chroma;
|
cannam@475
|
134 delete m_Decimator;
|
cannam@475
|
135
|
cannam@475
|
136 delete [] m_DecimatedBuffer;
|
cannam@475
|
137 delete [] m_ChromaBuffer;
|
cannam@475
|
138 delete [] m_MeanHPCP;
|
cannam@475
|
139 delete [] m_MajCorr;
|
cannam@475
|
140 delete [] m_MinCorr;
|
cannam@475
|
141 delete [] m_MajProfileNorm;
|
cannam@475
|
142 delete [] m_MinProfileNorm;
|
cannam@475
|
143 delete [] m_MedianFilterBuffer;
|
cannam@475
|
144 delete [] m_SortedBuffer;
|
cannam@475
|
145 delete [] m_keyStrengths;
|
cannam@475
|
146 }
|
cannam@475
|
147
|
cannam@475
|
148 double GetKeyMode::krumCorr( const double *pDataNorm, const double *pProfileNorm,
|
cannam@475
|
149 int shiftProfile, unsigned int length)
|
cannam@475
|
150 {
|
cannam@475
|
151 double retVal= 0.0;
|
cannam@475
|
152
|
cannam@475
|
153 double num = 0;
|
cannam@475
|
154 double den = 0;
|
cannam@475
|
155 double sum1 = 0;
|
cannam@475
|
156 double sum2 = 0;
|
cannam@475
|
157
|
cannam@478
|
158 for( unsigned int i = 0; i <length; i++ ) {
|
cannam@478
|
159
|
cannam@475
|
160 int k = (i - shiftProfile + length) % length;
|
cannam@475
|
161
|
cannam@475
|
162 num += pDataNorm[i] * pProfileNorm[k];
|
cannam@475
|
163
|
cannam@475
|
164 sum1 += ( pDataNorm[i] * pDataNorm[i] );
|
cannam@475
|
165 sum2 += ( pProfileNorm[k] * pProfileNorm[k] );
|
cannam@475
|
166 }
|
cannam@478
|
167
|
cannam@475
|
168 den = sqrt(sum1 * sum2);
|
cannam@475
|
169
|
cannam@475
|
170 if( den>0 ) {
|
cannam@475
|
171 retVal = num/den;
|
cannam@475
|
172 } else {
|
cannam@475
|
173 retVal = 0;
|
cannam@475
|
174 }
|
cannam@475
|
175
|
cannam@475
|
176 return retVal;
|
cannam@475
|
177 }
|
cannam@475
|
178
|
cannam@475
|
179 int GetKeyMode::process(double *PCMData)
|
cannam@475
|
180 {
|
cannam@475
|
181 int key;
|
cannam@475
|
182 unsigned int j,k;
|
cannam@475
|
183
|
cannam@475
|
184 m_Decimator->process( PCMData, m_DecimatedBuffer);
|
cannam@475
|
185
|
cannam@475
|
186 m_ChrPointer = m_Chroma->process( m_DecimatedBuffer );
|
cannam@475
|
187
|
cannam@475
|
188 // populate hpcp values;
|
cannam@475
|
189 int cbidx;
|
cannam@475
|
190 for( j = 0; j < kBinsPerOctave; j++ ) {
|
cannam@475
|
191 cbidx = (m_bufferindex * kBinsPerOctave) + j;
|
cannam@475
|
192 m_ChromaBuffer[ cbidx ] = m_ChrPointer[j];
|
cannam@475
|
193 }
|
cannam@475
|
194
|
cannam@475
|
195 //keep track of input buffers;
|
cannam@475
|
196 if( m_bufferindex++ >= m_ChromaBuffersize - 1) {
|
cannam@475
|
197 m_bufferindex = 0;
|
cannam@475
|
198 }
|
cannam@475
|
199
|
cannam@475
|
200 // track filling of chroma matrix
|
cannam@475
|
201 if( m_ChromaBufferFilling++ >= m_ChromaBuffersize) {
|
cannam@475
|
202 m_ChromaBufferFilling = m_ChromaBuffersize;
|
cannam@475
|
203 }
|
cannam@475
|
204
|
cannam@475
|
205 //calculate mean
|
cannam@475
|
206 for( k = 0; k < kBinsPerOctave; k++ ) {
|
cannam@475
|
207 double mnVal = 0.0;
|
cannam@475
|
208 for( j = 0; j < m_ChromaBufferFilling; j++ ) {
|
cannam@475
|
209 mnVal += m_ChromaBuffer[ k + (j*kBinsPerOctave) ];
|
cannam@475
|
210 }
|
cannam@475
|
211
|
cannam@475
|
212 m_MeanHPCP[k] = mnVal/(double)m_ChromaBufferFilling;
|
cannam@475
|
213 }
|
cannam@475
|
214
|
cannam@475
|
215 // Normalize for zero average
|
cannam@475
|
216 double mHPCP = MathUtilities::mean( m_MeanHPCP, kBinsPerOctave );
|
cannam@478
|
217 for( k = 0; k < kBinsPerOctave; k++ ) {
|
cannam@475
|
218 m_MeanHPCP[k] -= mHPCP;
|
cannam@475
|
219 }
|
cannam@475
|
220
|
cannam@478
|
221 for( k = 0; k < kBinsPerOctave; k++ ) {
|
cannam@478
|
222 // The Chromagram has the center of C at bin 0, while the major
|
cannam@475
|
223 // and minor profiles have the center of C at 1. We want to have
|
cannam@475
|
224 // the correlation for C result also at 1.
|
cannam@475
|
225 // To achieve this we have to shift two times:
|
cannam@475
|
226 m_MajCorr[k] = krumCorr( m_MeanHPCP, m_MajProfileNorm, (int)k - 2, kBinsPerOctave );
|
cannam@475
|
227 m_MinCorr[k] = krumCorr( m_MeanHPCP, m_MinProfileNorm, (int)k - 2, kBinsPerOctave );
|
cannam@475
|
228 }
|
cannam@475
|
229
|
cannam@475
|
230 // m_MajCorr[1] is C center 1 / 3 + 1 = 1
|
cannam@475
|
231 // m_MajCorr[4] is D center 4 / 3 + 1 = 2
|
cannam@475
|
232 // '+ 1' because we number keys 1-24, not 0-23.
|
cannam@475
|
233 double maxMaj;
|
cannam@475
|
234 int maxMajBin = MathUtilities::getMax( m_MajCorr, kBinsPerOctave, &maxMaj );
|
cannam@475
|
235 double maxMin;
|
cannam@475
|
236 int maxMinBin = MathUtilities::getMax( m_MinCorr, kBinsPerOctave, &maxMin );
|
cannam@475
|
237 int maxBin = (maxMaj > maxMin) ? maxMajBin : (maxMinBin + kBinsPerOctave);
|
cannam@475
|
238 key = maxBin / 3 + 1;
|
cannam@475
|
239
|
cannam@475
|
240 //Median filtering
|
cannam@475
|
241
|
cannam@475
|
242 // track Median buffer initial filling
|
cannam@475
|
243 if( m_MedianBufferFilling++ >= m_MedianWinsize) {
|
cannam@475
|
244 m_MedianBufferFilling = m_MedianWinsize;
|
cannam@475
|
245 }
|
cannam@475
|
246
|
cannam@475
|
247 //shift median buffer
|
cannam@475
|
248 for( k = 1; k < m_MedianWinsize; k++ ) {
|
cannam@475
|
249 m_MedianFilterBuffer[ k - 1 ] = m_MedianFilterBuffer[ k ];
|
cannam@475
|
250 }
|
cannam@475
|
251
|
cannam@475
|
252 //write new key value into median buffer
|
cannam@475
|
253 m_MedianFilterBuffer[ m_MedianWinsize - 1 ] = key;
|
cannam@475
|
254
|
cannam@475
|
255
|
cannam@475
|
256 //Copy median into sorting buffer, reversed
|
cannam@475
|
257 unsigned int ijx = 0;
|
cannam@475
|
258 for( k = 0; k < m_MedianWinsize; k++ ) {
|
cannam@475
|
259 m_SortedBuffer[k] = m_MedianFilterBuffer[m_MedianWinsize-1-ijx];
|
cannam@475
|
260 ijx++;
|
cannam@475
|
261 }
|
cannam@475
|
262
|
cannam@475
|
263 qsort(m_SortedBuffer, m_MedianBufferFilling, sizeof(unsigned int),
|
cannam@475
|
264 MathUtilities::compareInt);
|
cannam@478
|
265
|
cannam@475
|
266 int sortlength = m_MedianBufferFilling;
|
cannam@475
|
267 int midpoint = (int)ceil((double)sortlength/2);
|
cannam@475
|
268
|
cannam@475
|
269 if( midpoint <= 0 ) {
|
cannam@475
|
270 midpoint = 1;
|
cannam@475
|
271 }
|
cannam@475
|
272
|
cannam@475
|
273 key = m_SortedBuffer[midpoint-1];
|
cannam@475
|
274
|
cannam@475
|
275 return key;
|
cannam@475
|
276 }
|
cannam@475
|
277
|
cannam@475
|
278
|
cannam@475
|
279 bool GetKeyMode::isModeMinor( int key )
|
cannam@475
|
280 {
|
cannam@475
|
281 return (key > 12);
|
cannam@475
|
282 }
|
cannam@475
|
283
|
cannam@475
|
284 unsigned int getChromaSize()
|
cannam@475
|
285 {
|
cannam@475
|
286 return kBinsPerOctave;
|
cannam@475
|
287 }
|
cannam@475
|
288
|
cannam@475
|
289 double* GetKeyMode::getKeyStrengths() {
|
cannam@475
|
290 unsigned int k;
|
cannam@475
|
291
|
cannam@475
|
292 for (k = 0; k < 24; ++k) {
|
cannam@475
|
293 m_keyStrengths[k] = 0;
|
cannam@475
|
294 }
|
cannam@475
|
295
|
cannam@478
|
296 for( k = 0; k < kBinsPerOctave; k++ ) {
|
cannam@475
|
297 int idx = k / (kBinsPerOctave/12);
|
cannam@475
|
298 int rem = k % (kBinsPerOctave/12);
|
cannam@475
|
299 if (rem == 0 || m_MajCorr[k] > m_keyStrengths[idx]) {
|
cannam@475
|
300 m_keyStrengths[idx] = m_MajCorr[k];
|
cannam@475
|
301 }
|
cannam@475
|
302 }
|
cannam@475
|
303
|
cannam@478
|
304 for( k = 0; k < kBinsPerOctave; k++ ) {
|
cannam@475
|
305 int idx = (k + kBinsPerOctave) / (kBinsPerOctave/12);
|
cannam@475
|
306 int rem = k % (kBinsPerOctave/12);
|
cannam@475
|
307 if (rem == 0 || m_MinCorr[k] > m_keyStrengths[idx]) {
|
cannam@475
|
308 m_keyStrengths[idx] = m_MinCorr[k];
|
cannam@475
|
309 }
|
cannam@475
|
310 }
|
cannam@475
|
311
|
cannam@475
|
312 return m_keyStrengths;
|
cannam@475
|
313 }
|