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
|
c@232
|
25 // Chords profile
|
daschuer@457
|
26 static double MajProfile[36] = {
|
daschuer@457
|
27 0.0384, 0.0629, 0.0258, 0.0121, 0.0146, 0.0106, 0.0364, 0.0610, 0.0267,
|
daschuer@457
|
28 0.0126, 0.0121, 0.0086, 0.0364, 0.0623, 0.0279, 0.0275, 0.0414, 0.0186,
|
daschuer@457
|
29 0.0173, 0.0248, 0.0145, 0.0364, 0.0631, 0.0262, 0.0129, 0.0150, 0.0098,
|
daschuer@457
|
30 0.0312, 0.0521, 0.0235, 0.0129, 0.0142, 0.0095, 0.0289, 0.0478, 0.0239};
|
c@232
|
31
|
daschuer@457
|
32 static double MinProfile[36] = {
|
daschuer@457
|
33 0.0375, 0.0682, 0.0299, 0.0119, 0.0138, 0.0093, 0.0296, 0.0543, 0.0257,
|
daschuer@457
|
34 0.0292, 0.0519, 0.0246, 0.0159, 0.0234, 0.0135, 0.0291, 0.0544, 0.0248,
|
daschuer@457
|
35 0.0137, 0.0176, 0.0104, 0.0352, 0.0670, 0.0302, 0.0222, 0.0349, 0.0164,
|
daschuer@457
|
36 0.0174, 0.0297, 0.0166, 0.0222, 0.0401, 0.0202, 0.0175, 0.0270, 0.0146};
|
c@232
|
37 //
|
c@265
|
38
|
c@232
|
39
|
c@232
|
40 //////////////////////////////////////////////////////////////////////
|
c@232
|
41 // Construction/Destruction
|
c@232
|
42 //////////////////////////////////////////////////////////////////////
|
c@232
|
43
|
c@234
|
44 GetKeyMode::GetKeyMode( int sampleRate, float tuningFrequency,
|
daschuer@457
|
45 double hpcpAverage, double medianAverage ) :
|
c@259
|
46 m_hpcpAverage( hpcpAverage ),
|
c@259
|
47 m_medianAverage( medianAverage ),
|
c@259
|
48 m_ChrPointer(0),
|
c@259
|
49 m_DecimatedBuffer(0),
|
c@259
|
50 m_ChromaBuffer(0),
|
c@259
|
51 m_MeanHPCP(0),
|
c@259
|
52 m_MajCorr(0),
|
c@259
|
53 m_MinCorr(0),
|
c@259
|
54 m_Keys(0),
|
c@259
|
55 m_MedianFilterBuffer(0),
|
c@265
|
56 m_SortedBuffer(0),
|
c@265
|
57 m_keyStrengths(0)
|
c@232
|
58 {
|
c@259
|
59 m_DecimationFactor = 8;
|
c@259
|
60
|
c@259
|
61 // Chromagram configuration parameters
|
c@259
|
62 m_ChromaConfig.normalise = MathUtilities::NormaliseUnitMax;
|
c@259
|
63 m_ChromaConfig.FS = lrint(sampleRate/(double)m_DecimationFactor);
|
daschuer@457
|
64 if (m_ChromaConfig.FS < 1) {
|
daschuer@457
|
65 m_ChromaConfig.FS = 1;
|
daschuer@457
|
66 }
|
c@232
|
67
|
c@259
|
68 // Set C (= MIDI #12) as our base :
|
c@259
|
69 // This implies that key = 1 => Cmaj, key = 12 => Bmaj, key = 13 => Cmin, etc.
|
daschuer@457
|
70 m_ChromaConfig.min = Pitch::getFrequencyForPitch( 48, 0, tuningFrequency );
|
daschuer@457
|
71 m_ChromaConfig.max = Pitch::getFrequencyForPitch( 96, 0, tuningFrequency );
|
c@234
|
72
|
c@259
|
73 m_ChromaConfig.BPO = 36;
|
c@259
|
74 m_ChromaConfig.CQThresh = 0.0054;
|
c@234
|
75
|
c@259
|
76 // Chromagram inst.
|
c@259
|
77 m_Chroma = new Chromagram( m_ChromaConfig );
|
c@235
|
78
|
c@259
|
79 // Get calculated parameters from chroma object
|
c@259
|
80 m_ChromaFrameSize = m_Chroma->getFrameSize();
|
c@259
|
81 // override hopsize for this application
|
c@259
|
82 m_ChromaHopSize = m_ChromaFrameSize;
|
c@259
|
83 m_BPO = m_ChromaConfig.BPO;
|
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
|
c@259
|
99 m_ChromaBuffer = new double[m_BPO * m_ChromaBuffersize];
|
c@259
|
100 memset( m_ChromaBuffer, 0, sizeof(double) * m_BPO * m_ChromaBuffersize);
|
c@259
|
101
|
c@259
|
102 m_MeanHPCP = new double[m_BPO];
|
c@259
|
103
|
c@259
|
104 m_MajCorr = new double[m_BPO];
|
c@259
|
105 m_MinCorr = new double[m_BPO];
|
c@259
|
106 m_Keys = new double[2*m_BPO];
|
c@259
|
107
|
c@259
|
108 m_MedianFilterBuffer = new int[ m_MedianWinsize ];
|
c@259
|
109 memset( m_MedianFilterBuffer, 0, sizeof(int)*m_MedianWinsize);
|
c@259
|
110
|
c@259
|
111 m_SortedBuffer = new int[ m_MedianWinsize ];
|
daschuer@457
|
112 memset( m_SortedBuffer, 0, sizeof(int)*m_MedianWinsize);
|
c@259
|
113
|
daschuer@457
|
114 m_Decimator = new Decimator( m_ChromaFrameSize*m_DecimationFactor, m_DecimationFactor );
|
c@265
|
115
|
c@265
|
116 m_keyStrengths = new double[24];
|
c@232
|
117 }
|
c@232
|
118
|
c@232
|
119 GetKeyMode::~GetKeyMode()
|
c@232
|
120 {
|
c@259
|
121 delete m_Chroma;
|
c@259
|
122 delete m_Decimator;
|
c@259
|
123
|
c@259
|
124 delete [] m_DecimatedBuffer;
|
c@259
|
125 delete [] m_ChromaBuffer;
|
c@259
|
126 delete [] m_MeanHPCP;
|
c@259
|
127 delete [] m_MajCorr;
|
c@259
|
128 delete [] m_MinCorr;
|
c@259
|
129 delete [] m_Keys;
|
c@259
|
130 delete [] m_MedianFilterBuffer;
|
c@259
|
131 delete [] m_SortedBuffer;
|
c@265
|
132
|
c@265
|
133 delete[] m_keyStrengths;
|
c@232
|
134 }
|
c@232
|
135
|
c@232
|
136 double GetKeyMode::krumCorr(double *pData1, double *pData2, unsigned int length)
|
c@232
|
137 {
|
c@259
|
138 double retVal= 0.0;
|
c@259
|
139
|
c@259
|
140 double num = 0;
|
c@259
|
141 double den = 0;
|
c@259
|
142 double mX = MathUtilities::mean( pData1, length );
|
c@259
|
143 double mY = MathUtilities::mean( pData2, length );
|
c@259
|
144
|
c@259
|
145 double sum1 = 0;
|
c@259
|
146 double sum2 = 0;
|
c@259
|
147
|
daschuer@457
|
148 for( unsigned int i = 0; i <length; i++ ) {
|
c@259
|
149 num += ( pData1[i] - mX ) * ( pData2[i] - mY );
|
c@232
|
150
|
c@259
|
151 sum1 += ( (pData1[i]-mX) * (pData1[i]-mX) );
|
c@259
|
152 sum2 += ( (pData2[i]-mY) * (pData2[i]-mY) );
|
c@259
|
153 }
|
daschuer@457
|
154
|
c@259
|
155 den = sqrt(sum1 * sum2);
|
daschuer@457
|
156
|
daschuer@457
|
157 if( den>0 ) {
|
c@259
|
158 retVal = num/den;
|
daschuer@457
|
159 } else {
|
c@259
|
160 retVal = 0;
|
daschuer@457
|
161 }
|
c@232
|
162
|
c@259
|
163 return retVal;
|
c@232
|
164 }
|
c@232
|
165
|
c@232
|
166 int GetKeyMode::process(double *PCMData)
|
c@232
|
167 {
|
c@259
|
168 int key;
|
c@259
|
169 unsigned int j,k;
|
c@232
|
170
|
c@259
|
171 //////////////////////////////////////////////
|
c@259
|
172 m_Decimator->process( PCMData, m_DecimatedBuffer);
|
c@232
|
173
|
daschuer@457
|
174 m_ChrPointer = m_Chroma->process( m_DecimatedBuffer );
|
c@232
|
175
|
daschuer@456
|
176 // The Cromagram has the center of C at bin 0, while the major
|
daschuer@456
|
177 // and minor profiles have the center of C at 1. We want to have
|
daschuer@456
|
178 // the correlation for C result also at 1.
|
daschuer@456
|
179 // To achieve this we have to shift two times:
|
daschuer@456
|
180 MathUtilities::circShift( m_ChrPointer, m_BPO, 2);
|
c@248
|
181 /*
|
c@298
|
182 std::cout << "raw chroma: ";
|
c@298
|
183 for (int ii = 0; ii < m_BPO; ++ii) {
|
c@298
|
184 if (ii % (m_BPO/12) == 0) std::cout << "\n";
|
c@298
|
185 std::cout << m_ChrPointer[ii] << " ";
|
c@298
|
186 }
|
c@298
|
187 std::cout << std::endl;
|
c@248
|
188 */
|
c@259
|
189 // populate hpcp values;
|
c@259
|
190 int cbidx;
|
daschuer@457
|
191 for( j = 0; j < m_BPO; j++ ) {
|
c@259
|
192 cbidx = (m_bufferindex * m_BPO) + j;
|
c@259
|
193 m_ChromaBuffer[ cbidx ] = m_ChrPointer[j];
|
c@259
|
194 }
|
c@232
|
195
|
c@259
|
196 //keep track of input buffers;
|
daschuer@457
|
197 if( m_bufferindex++ >= m_ChromaBuffersize - 1) {
|
c@259
|
198 m_bufferindex = 0;
|
daschuer@457
|
199 }
|
c@232
|
200
|
c@259
|
201 // track filling of chroma matrix
|
daschuer@457
|
202 if( m_ChromaBufferFilling++ >= m_ChromaBuffersize) {
|
c@259
|
203 m_ChromaBufferFilling = m_ChromaBuffersize;
|
daschuer@457
|
204 }
|
c@232
|
205
|
daschuer@457
|
206 //calculate mean
|
daschuer@457
|
207 for( k = 0; k < m_BPO; k++ ) {
|
c@259
|
208 double mnVal = 0.0;
|
daschuer@457
|
209 for( j = 0; j < m_ChromaBufferFilling; j++ ) {
|
c@259
|
210 mnVal += m_ChromaBuffer[ k + (j*m_BPO) ];
|
c@259
|
211 }
|
c@232
|
212
|
c@259
|
213 m_MeanHPCP[k] = mnVal/(double)m_ChromaBufferFilling;
|
c@259
|
214 }
|
c@232
|
215
|
c@232
|
216
|
daschuer@457
|
217 for( k = 0; k < m_BPO; k++ ) {
|
c@259
|
218 m_MajCorr[k] = krumCorr( m_MeanHPCP, MajProfile, m_BPO );
|
c@259
|
219 m_MinCorr[k] = krumCorr( m_MeanHPCP, MinProfile, m_BPO );
|
c@232
|
220
|
c@259
|
221 MathUtilities::circShift( MajProfile, m_BPO, 1 );
|
c@259
|
222 MathUtilities::circShift( MinProfile, m_BPO, 1 );
|
c@259
|
223 }
|
daschuer@457
|
224
|
daschuer@457
|
225 for( k = 0; k < m_BPO; k++ ) {
|
c@259
|
226 m_Keys[k] = m_MajCorr[k];
|
c@259
|
227 m_Keys[k+m_BPO] = m_MinCorr[k];
|
c@259
|
228 }
|
devnull@240
|
229
|
c@265
|
230 for (k = 0; k < 24; ++k) {
|
c@265
|
231 m_keyStrengths[k] = 0;
|
c@265
|
232 }
|
c@265
|
233
|
daschuer@457
|
234 for( k = 0; k < m_BPO*2; k++ ) {
|
c@298
|
235 int idx = k / (m_BPO/12);
|
c@298
|
236 int rem = k % (m_BPO/12);
|
c@298
|
237 if (rem == 0 || m_Keys[k] > m_keyStrengths[idx]) {
|
c@298
|
238 m_keyStrengths[idx] = m_Keys[k];
|
c@298
|
239 }
|
c@298
|
240
|
c@298
|
241 // m_keyStrengths[k/(m_BPO/12)] += m_Keys[k];
|
c@265
|
242 }
|
devnull@240
|
243
|
c@248
|
244 /*
|
c@259
|
245 std::cout << "raw keys: ";
|
c@259
|
246 for (int ii = 0; ii < 2*m_BPO; ++ii) {
|
c@298
|
247 if (ii % (m_BPO/12) == 0) std::cout << "\n";
|
c@298
|
248 std::cout << m_Keys[ii] << " ";
|
c@298
|
249 }
|
c@298
|
250 std::cout << std::endl;
|
c@298
|
251
|
c@298
|
252 std::cout << "key strengths: ";
|
c@298
|
253 for (int ii = 0; ii < 24; ++ii) {
|
c@298
|
254 if (ii % 6 == 0) std::cout << "\n";
|
c@298
|
255 std::cout << m_keyStrengths[ii] << " ";
|
c@259
|
256 }
|
c@259
|
257 std::cout << std::endl;
|
c@248
|
258 */
|
c@259
|
259 double dummy;
|
daschuer@456
|
260 // m_Keys[1] is C center 1 / 3 + 1 = 1
|
daschuer@456
|
261 // m_Keys[4] is D center 4 / 3 + 1 = 2
|
daschuer@456
|
262 // '+ 1' because we number keys 1-24, not 0-23.
|
daschuer@456
|
263 key = MathUtilities::getMax( m_Keys, 2* m_BPO, &dummy ) / 3 + 1;
|
c@234
|
264
|
c@298
|
265 // std::cout << "key pre-sorting: " << key << std::endl;
|
c@232
|
266
|
c@232
|
267
|
c@259
|
268 //Median filtering
|
c@232
|
269
|
c@259
|
270 // track Median buffer initial filling
|
daschuer@457
|
271 if( m_MedianBufferFilling++ >= m_MedianWinsize) {
|
c@259
|
272 m_MedianBufferFilling = m_MedianWinsize;
|
daschuer@457
|
273 }
|
daschuer@457
|
274
|
c@259
|
275 //shift median buffer
|
daschuer@457
|
276 for( k = 1; k < m_MedianWinsize; k++ ) {
|
c@259
|
277 m_MedianFilterBuffer[ k - 1 ] = m_MedianFilterBuffer[ k ];
|
c@259
|
278 }
|
c@232
|
279
|
c@259
|
280 //write new key value into median buffer
|
c@259
|
281 m_MedianFilterBuffer[ m_MedianWinsize - 1 ] = key;
|
c@232
|
282
|
c@232
|
283
|
c@259
|
284 //Copy median into sorting buffer, reversed
|
c@259
|
285 unsigned int ijx = 0;
|
daschuer@457
|
286 for( k = 0; k < m_MedianWinsize; k++ ) {
|
c@259
|
287 m_SortedBuffer[k] = m_MedianFilterBuffer[m_MedianWinsize-1-ijx];
|
c@259
|
288 ijx++;
|
c@259
|
289 }
|
c@232
|
290
|
c@259
|
291 qsort(m_SortedBuffer, m_MedianBufferFilling, sizeof(unsigned int),
|
c@259
|
292 MathUtilities::compareInt);
|
c@259
|
293 /*
|
c@259
|
294 std::cout << "sorted: ";
|
c@259
|
295 for (int ii = 0; ii < m_MedianBufferFilling; ++ii) {
|
c@259
|
296 std::cout << m_SortedBuffer[ii] << " ";
|
c@259
|
297 }
|
c@259
|
298 std::cout << std::endl;
|
c@259
|
299 */
|
c@259
|
300 int sortlength = m_MedianBufferFilling;
|
c@259
|
301 int midpoint = (int)ceil((double)sortlength/2);
|
c@232
|
302
|
c@298
|
303 // std::cout << "midpoint = " << midpoint << endl;
|
c@298
|
304
|
daschuer@457
|
305 if( midpoint <= 0 ) {
|
c@259
|
306 midpoint = 1;
|
daschuer@457
|
307 }
|
c@232
|
308
|
c@259
|
309 key = m_SortedBuffer[midpoint-1];
|
c@232
|
310
|
c@298
|
311 // std::cout << "returning key = " << key << endl;
|
c@298
|
312
|
c@259
|
313 return key;
|
c@232
|
314 }
|
c@232
|
315
|
c@232
|
316
|
c@268
|
317 bool GetKeyMode::isModeMinor( int key )
|
c@232
|
318 {
|
c@259
|
319 return (key > 12);
|
c@232
|
320 }
|