c@259
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
c@378
|
2 /*
|
c@378
|
3 Copyright (c) 2005 Centre for Digital Music ( C4DM )
|
c@378
|
4 Queen Mary Univesrity of London
|
c@259
|
5
|
c@378
|
6 This program is free software; you can redistribute it and/or
|
c@378
|
7 modify it under the terms of the GNU General Public License as
|
c@378
|
8 published by the Free Software Foundation; either version 2 of the
|
c@378
|
9 License, or (at your option) any later version. See the file
|
c@378
|
10 COPYING included with this distribution for more information.
|
c@378
|
11 */
|
c@232
|
12 // GetKeyMode.cpp: implementation of the CGetKeyMode class.
|
c@232
|
13 //
|
c@232
|
14 //////////////////////////////////////////////////////////////////////
|
c@232
|
15
|
c@232
|
16 #include "GetKeyMode.h"
|
c@241
|
17 #include "maths/MathUtilities.h"
|
c@234
|
18 #include "base/Pitch.h"
|
c@234
|
19
|
c@234
|
20 #include <iostream>
|
c@232
|
21
|
c@272
|
22 #include <cstring>
|
c@272
|
23 #include <cstdlib>
|
c@272
|
24
|
daschuer@458
|
25 static const int kBinsPerOctave = 36;
|
daschuer@458
|
26
|
c@232
|
27 // Chords profile
|
daschuer@458
|
28 static double MajProfile[kBinsPerOctave] = {
|
daschuer@457
|
29 0.0384, 0.0629, 0.0258, 0.0121, 0.0146, 0.0106, 0.0364, 0.0610, 0.0267,
|
daschuer@457
|
30 0.0126, 0.0121, 0.0086, 0.0364, 0.0623, 0.0279, 0.0275, 0.0414, 0.0186,
|
daschuer@457
|
31 0.0173, 0.0248, 0.0145, 0.0364, 0.0631, 0.0262, 0.0129, 0.0150, 0.0098,
|
daschuer@457
|
32 0.0312, 0.0521, 0.0235, 0.0129, 0.0142, 0.0095, 0.0289, 0.0478, 0.0239};
|
c@232
|
33
|
daschuer@458
|
34 static double MinProfile[kBinsPerOctave] = {
|
daschuer@457
|
35 0.0375, 0.0682, 0.0299, 0.0119, 0.0138, 0.0093, 0.0296, 0.0543, 0.0257,
|
daschuer@457
|
36 0.0292, 0.0519, 0.0246, 0.0159, 0.0234, 0.0135, 0.0291, 0.0544, 0.0248,
|
daschuer@457
|
37 0.0137, 0.0176, 0.0104, 0.0352, 0.0670, 0.0302, 0.0222, 0.0349, 0.0164,
|
daschuer@457
|
38 0.0174, 0.0297, 0.0166, 0.0222, 0.0401, 0.0202, 0.0175, 0.0270, 0.0146};
|
c@232
|
39 //
|
c@265
|
40
|
c@232
|
41
|
c@232
|
42 //////////////////////////////////////////////////////////////////////
|
c@232
|
43 // Construction/Destruction
|
c@232
|
44 //////////////////////////////////////////////////////////////////////
|
c@232
|
45
|
c@234
|
46 GetKeyMode::GetKeyMode( int sampleRate, float tuningFrequency,
|
daschuer@457
|
47 double hpcpAverage, double medianAverage ) :
|
c@259
|
48 m_hpcpAverage( hpcpAverage ),
|
c@259
|
49 m_medianAverage( medianAverage ),
|
c@259
|
50 m_ChrPointer(0),
|
c@259
|
51 m_DecimatedBuffer(0),
|
c@259
|
52 m_ChromaBuffer(0),
|
c@259
|
53 m_MeanHPCP(0),
|
c@259
|
54 m_MajCorr(0),
|
c@259
|
55 m_MinCorr(0),
|
c@259
|
56 m_MedianFilterBuffer(0),
|
c@265
|
57 m_SortedBuffer(0),
|
c@265
|
58 m_keyStrengths(0)
|
c@232
|
59 {
|
c@259
|
60 m_DecimationFactor = 8;
|
c@259
|
61
|
c@259
|
62 // Chromagram configuration parameters
|
c@259
|
63 m_ChromaConfig.normalise = MathUtilities::NormaliseUnitMax;
|
c@259
|
64 m_ChromaConfig.FS = lrint(sampleRate/(double)m_DecimationFactor);
|
daschuer@457
|
65 if (m_ChromaConfig.FS < 1) {
|
daschuer@457
|
66 m_ChromaConfig.FS = 1;
|
daschuer@457
|
67 }
|
c@232
|
68
|
c@259
|
69 // Set C (= MIDI #12) as our base :
|
c@259
|
70 // This implies that key = 1 => Cmaj, key = 12 => Bmaj, key = 13 => Cmin, etc.
|
daschuer@457
|
71 m_ChromaConfig.min = Pitch::getFrequencyForPitch( 48, 0, tuningFrequency );
|
daschuer@457
|
72 m_ChromaConfig.max = Pitch::getFrequencyForPitch( 96, 0, tuningFrequency );
|
c@234
|
73
|
daschuer@458
|
74 m_ChromaConfig.BPO = kBinsPerOctave;
|
c@259
|
75 m_ChromaConfig.CQThresh = 0.0054;
|
c@234
|
76
|
c@259
|
77 // Chromagram inst.
|
c@259
|
78 m_Chroma = new Chromagram( m_ChromaConfig );
|
c@235
|
79
|
c@259
|
80 // Get calculated parameters from chroma object
|
c@259
|
81 m_ChromaFrameSize = m_Chroma->getFrameSize();
|
c@259
|
82 // override hopsize for this application
|
c@259
|
83 m_ChromaHopSize = m_ChromaFrameSize;
|
c@235
|
84
|
c@265
|
85 // std::cerr << "chroma frame size = " << m_ChromaFrameSize << ", decimation factor = " << m_DecimationFactor << " therefore block size = " << getBlockSize() << std::endl;
|
c@265
|
86
|
c@259
|
87 // Chromagram average and estimated key median filter lengths
|
c@259
|
88 m_ChromaBuffersize = (int)ceil( m_hpcpAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
|
c@259
|
89 m_MedianWinsize = (int)ceil( m_medianAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
|
c@259
|
90
|
c@259
|
91 // Reset counters
|
c@259
|
92 m_bufferindex = 0;
|
c@259
|
93 m_ChromaBufferFilling = 0;
|
c@259
|
94 m_MedianBufferFilling = 0;
|
c@234
|
95
|
c@259
|
96 // Spawn objectc/arrays
|
c@259
|
97 m_DecimatedBuffer = new double[m_ChromaFrameSize];
|
c@259
|
98
|
daschuer@458
|
99 m_ChromaBuffer = new double[kBinsPerOctave * m_ChromaBuffersize];
|
daschuer@458
|
100 memset( m_ChromaBuffer, 0, sizeof(double) * kBinsPerOctave * m_ChromaBuffersize);
|
c@259
|
101
|
daschuer@458
|
102 m_MeanHPCP = new double[kBinsPerOctave];
|
c@259
|
103
|
daschuer@458
|
104 m_MajCorr = new double[kBinsPerOctave];
|
daschuer@458
|
105 m_MinCorr = new double[kBinsPerOctave];
|
c@259
|
106
|
c@259
|
107 m_MedianFilterBuffer = new int[ m_MedianWinsize ];
|
c@259
|
108 memset( m_MedianFilterBuffer, 0, sizeof(int)*m_MedianWinsize);
|
c@259
|
109
|
c@259
|
110 m_SortedBuffer = new int[ m_MedianWinsize ];
|
daschuer@457
|
111 memset( m_SortedBuffer, 0, sizeof(int)*m_MedianWinsize);
|
c@259
|
112
|
daschuer@457
|
113 m_Decimator = new Decimator( m_ChromaFrameSize*m_DecimationFactor, m_DecimationFactor );
|
c@265
|
114
|
c@265
|
115 m_keyStrengths = new double[24];
|
c@232
|
116 }
|
c@232
|
117
|
c@232
|
118 GetKeyMode::~GetKeyMode()
|
c@232
|
119 {
|
c@259
|
120 delete m_Chroma;
|
c@259
|
121 delete m_Decimator;
|
c@259
|
122
|
c@259
|
123 delete [] m_DecimatedBuffer;
|
c@259
|
124 delete [] m_ChromaBuffer;
|
c@259
|
125 delete [] m_MeanHPCP;
|
c@259
|
126 delete [] m_MajCorr;
|
c@259
|
127 delete [] m_MinCorr;
|
c@259
|
128 delete [] m_MedianFilterBuffer;
|
c@259
|
129 delete [] m_SortedBuffer;
|
daschuer@460
|
130 delete [] m_keyStrengths;
|
c@232
|
131 }
|
c@232
|
132
|
c@232
|
133 double GetKeyMode::krumCorr(double *pData1, double *pData2, unsigned int length)
|
c@232
|
134 {
|
c@259
|
135 double retVal= 0.0;
|
c@259
|
136
|
c@259
|
137 double num = 0;
|
c@259
|
138 double den = 0;
|
c@259
|
139 double mX = MathUtilities::mean( pData1, length );
|
c@259
|
140 double mY = MathUtilities::mean( pData2, length );
|
c@259
|
141
|
c@259
|
142 double sum1 = 0;
|
c@259
|
143 double sum2 = 0;
|
c@259
|
144
|
daschuer@457
|
145 for( unsigned int i = 0; i <length; i++ ) {
|
c@259
|
146 num += ( pData1[i] - mX ) * ( pData2[i] - mY );
|
c@232
|
147
|
c@259
|
148 sum1 += ( (pData1[i]-mX) * (pData1[i]-mX) );
|
c@259
|
149 sum2 += ( (pData2[i]-mY) * (pData2[i]-mY) );
|
c@259
|
150 }
|
daschuer@457
|
151
|
c@259
|
152 den = sqrt(sum1 * sum2);
|
daschuer@457
|
153
|
daschuer@457
|
154 if( den>0 ) {
|
c@259
|
155 retVal = num/den;
|
daschuer@457
|
156 } else {
|
c@259
|
157 retVal = 0;
|
daschuer@457
|
158 }
|
c@232
|
159
|
c@259
|
160 return retVal;
|
c@232
|
161 }
|
c@232
|
162
|
c@232
|
163 int GetKeyMode::process(double *PCMData)
|
c@232
|
164 {
|
c@259
|
165 int key;
|
c@259
|
166 unsigned int j,k;
|
c@232
|
167
|
c@259
|
168 //////////////////////////////////////////////
|
c@259
|
169 m_Decimator->process( PCMData, m_DecimatedBuffer);
|
c@232
|
170
|
daschuer@457
|
171 m_ChrPointer = m_Chroma->process( m_DecimatedBuffer );
|
c@232
|
172
|
daschuer@456
|
173 // The Cromagram has the center of C at bin 0, while the major
|
daschuer@456
|
174 // and minor profiles have the center of C at 1. We want to have
|
daschuer@456
|
175 // the correlation for C result also at 1.
|
daschuer@456
|
176 // To achieve this we have to shift two times:
|
daschuer@458
|
177 MathUtilities::circShift( m_ChrPointer, kBinsPerOctave, 2);
|
c@248
|
178 /*
|
c@298
|
179 std::cout << "raw chroma: ";
|
daschuer@458
|
180 for (int ii = 0; ii < kBinsPerOctave; ++ii) {
|
daschuer@458
|
181 if (ii % (kBinsPerOctave/12) == 0) std::cout << "\n";
|
c@298
|
182 std::cout << m_ChrPointer[ii] << " ";
|
c@298
|
183 }
|
c@298
|
184 std::cout << std::endl;
|
c@248
|
185 */
|
c@259
|
186 // populate hpcp values;
|
c@259
|
187 int cbidx;
|
daschuer@458
|
188 for( j = 0; j < kBinsPerOctave; j++ ) {
|
daschuer@458
|
189 cbidx = (m_bufferindex * kBinsPerOctave) + j;
|
c@259
|
190 m_ChromaBuffer[ cbidx ] = m_ChrPointer[j];
|
c@259
|
191 }
|
c@232
|
192
|
c@259
|
193 //keep track of input buffers;
|
daschuer@457
|
194 if( m_bufferindex++ >= m_ChromaBuffersize - 1) {
|
c@259
|
195 m_bufferindex = 0;
|
daschuer@457
|
196 }
|
c@232
|
197
|
c@259
|
198 // track filling of chroma matrix
|
daschuer@457
|
199 if( m_ChromaBufferFilling++ >= m_ChromaBuffersize) {
|
c@259
|
200 m_ChromaBufferFilling = m_ChromaBuffersize;
|
daschuer@457
|
201 }
|
c@232
|
202
|
daschuer@457
|
203 //calculate mean
|
daschuer@458
|
204 for( k = 0; k < kBinsPerOctave; k++ ) {
|
c@259
|
205 double mnVal = 0.0;
|
daschuer@457
|
206 for( j = 0; j < m_ChromaBufferFilling; j++ ) {
|
daschuer@458
|
207 mnVal += m_ChromaBuffer[ k + (j*kBinsPerOctave) ];
|
c@259
|
208 }
|
c@232
|
209
|
c@259
|
210 m_MeanHPCP[k] = mnVal/(double)m_ChromaBufferFilling;
|
c@259
|
211 }
|
c@232
|
212
|
c@232
|
213
|
daschuer@458
|
214 for( k = 0; k < kBinsPerOctave; k++ ) {
|
daschuer@458
|
215 m_MajCorr[k] = krumCorr( m_MeanHPCP, MajProfile, kBinsPerOctave );
|
daschuer@458
|
216 m_MinCorr[k] = krumCorr( m_MeanHPCP, MinProfile, kBinsPerOctave );
|
c@232
|
217
|
daschuer@458
|
218 MathUtilities::circShift( MajProfile, kBinsPerOctave, 1 );
|
daschuer@458
|
219 MathUtilities::circShift( MinProfile, kBinsPerOctave, 1 );
|
c@259
|
220 }
|
daschuer@457
|
221
|
c@248
|
222 /*
|
c@259
|
223 std::cout << "raw keys: ";
|
daschuer@460
|
224 for (int ii = 0; ii < kBinsPerOctave; ++ii) {
|
daschuer@458
|
225 if (ii % (kBinsPerOctave/12) == 0) std::cout << "\n";
|
daschuer@460
|
226 std::cout << m_MajCorr[ii] << " ";
|
daschuer@460
|
227 }
|
daschuer@460
|
228 for (int ii = 0; ii < kBinsPerOctave; ++ii) {
|
daschuer@460
|
229 if (ii % (kBinsPerOctave/12) == 0) std::cout << "\n";
|
daschuer@460
|
230 std::cout << m_MinCorr[ii] << " ";
|
c@298
|
231 }
|
c@298
|
232 std::cout << std::endl;
|
daschuer@459
|
233 */
|
c@298
|
234
|
daschuer@460
|
235 // m_MajCorr[1] is C center 1 / 3 + 1 = 1
|
daschuer@460
|
236 // m_MajCorr[4] is D center 4 / 3 + 1 = 2
|
daschuer@456
|
237 // '+ 1' because we number keys 1-24, not 0-23.
|
daschuer@460
|
238 double maxMaj;
|
daschuer@460
|
239 int maxMajBin = MathUtilities::getMax( m_MajCorr, kBinsPerOctave, &maxMaj );
|
daschuer@460
|
240 double maxMin;
|
daschuer@460
|
241 int maxMinBin = MathUtilities::getMax( m_MinCorr, kBinsPerOctave, &maxMin );
|
daschuer@460
|
242 int maxBin = (maxMaj > maxMin) ? maxMajBin : (maxMinBin + kBinsPerOctave);
|
daschuer@460
|
243 key = maxBin / 3 + 1;
|
c@234
|
244
|
daschuer@460
|
245 // std::cout << "fractional key pre-sorting: " << (maxBin + 2) / 3.0 << std::endl;
|
c@298
|
246 // std::cout << "key pre-sorting: " << key << std::endl;
|
c@232
|
247
|
c@232
|
248
|
c@259
|
249 //Median filtering
|
c@232
|
250
|
c@259
|
251 // track Median buffer initial filling
|
daschuer@457
|
252 if( m_MedianBufferFilling++ >= m_MedianWinsize) {
|
c@259
|
253 m_MedianBufferFilling = m_MedianWinsize;
|
daschuer@457
|
254 }
|
daschuer@457
|
255
|
c@259
|
256 //shift median buffer
|
daschuer@457
|
257 for( k = 1; k < m_MedianWinsize; k++ ) {
|
c@259
|
258 m_MedianFilterBuffer[ k - 1 ] = m_MedianFilterBuffer[ k ];
|
c@259
|
259 }
|
c@232
|
260
|
c@259
|
261 //write new key value into median buffer
|
c@259
|
262 m_MedianFilterBuffer[ m_MedianWinsize - 1 ] = key;
|
c@232
|
263
|
c@232
|
264
|
c@259
|
265 //Copy median into sorting buffer, reversed
|
c@259
|
266 unsigned int ijx = 0;
|
daschuer@457
|
267 for( k = 0; k < m_MedianWinsize; k++ ) {
|
c@259
|
268 m_SortedBuffer[k] = m_MedianFilterBuffer[m_MedianWinsize-1-ijx];
|
c@259
|
269 ijx++;
|
c@259
|
270 }
|
c@232
|
271
|
c@259
|
272 qsort(m_SortedBuffer, m_MedianBufferFilling, sizeof(unsigned int),
|
c@259
|
273 MathUtilities::compareInt);
|
c@259
|
274 /*
|
c@259
|
275 std::cout << "sorted: ";
|
c@259
|
276 for (int ii = 0; ii < m_MedianBufferFilling; ++ii) {
|
c@259
|
277 std::cout << m_SortedBuffer[ii] << " ";
|
c@259
|
278 }
|
c@259
|
279 std::cout << std::endl;
|
c@259
|
280 */
|
c@259
|
281 int sortlength = m_MedianBufferFilling;
|
c@259
|
282 int midpoint = (int)ceil((double)sortlength/2);
|
c@232
|
283
|
c@298
|
284 // std::cout << "midpoint = " << midpoint << endl;
|
c@298
|
285
|
daschuer@457
|
286 if( midpoint <= 0 ) {
|
c@259
|
287 midpoint = 1;
|
daschuer@457
|
288 }
|
c@232
|
289
|
c@259
|
290 key = m_SortedBuffer[midpoint-1];
|
c@232
|
291
|
c@298
|
292 // std::cout << "returning key = " << key << endl;
|
c@298
|
293
|
c@259
|
294 return key;
|
c@232
|
295 }
|
c@232
|
296
|
c@232
|
297
|
c@268
|
298 bool GetKeyMode::isModeMinor( int key )
|
c@232
|
299 {
|
c@259
|
300 return (key > 12);
|
c@232
|
301 }
|
daschuer@458
|
302
|
daschuer@458
|
303 unsigned int getChromaSize()
|
daschuer@458
|
304 {
|
daschuer@458
|
305 return kBinsPerOctave;
|
daschuer@458
|
306 }
|
daschuer@459
|
307
|
daschuer@459
|
308 double* GetKeyMode::getKeyStrengths() {
|
daschuer@459
|
309 unsigned int k;
|
daschuer@459
|
310
|
daschuer@459
|
311 for (k = 0; k < 24; ++k) {
|
daschuer@459
|
312 m_keyStrengths[k] = 0;
|
daschuer@459
|
313 }
|
daschuer@459
|
314
|
daschuer@459
|
315 for( k = 0; k < kBinsPerOctave; k++ )
|
daschuer@459
|
316 {
|
daschuer@459
|
317 int idx = k / (kBinsPerOctave/12);
|
daschuer@459
|
318 int rem = k % (kBinsPerOctave/12);
|
daschuer@459
|
319 if (rem == 0 || m_MajCorr[k] > m_keyStrengths[idx]) {
|
daschuer@459
|
320 m_keyStrengths[idx] = m_MajCorr[k];
|
daschuer@459
|
321 }
|
daschuer@459
|
322 }
|
daschuer@459
|
323
|
daschuer@459
|
324 for( k = 0; k < kBinsPerOctave; k++ )
|
daschuer@459
|
325 {
|
daschuer@459
|
326 int idx = (k + kBinsPerOctave) / (kBinsPerOctave/12);
|
daschuer@459
|
327 int rem = k % (kBinsPerOctave/12);
|
daschuer@459
|
328 if (rem == 0 || m_MinCorr[k] > m_keyStrengths[idx]) {
|
daschuer@459
|
329 m_keyStrengths[idx] = m_MinCorr[k];
|
daschuer@459
|
330 }
|
daschuer@459
|
331 }
|
daschuer@459
|
332
|
daschuer@459
|
333 /*
|
daschuer@459
|
334 std::cout << "key strengths: ";
|
daschuer@459
|
335 for (int ii = 0; ii < 24; ++ii) {
|
daschuer@459
|
336 if (ii % 6 == 0) std::cout << "\n";
|
daschuer@459
|
337 std::cout << m_keyStrengths[ii] << " ";
|
daschuer@459
|
338 }
|
daschuer@459
|
339 std::cout << std::endl;
|
daschuer@459
|
340 */
|
daschuer@459
|
341
|
daschuer@459
|
342 return m_keyStrengths;
|
daschuer@459
|
343 }
|