cannam@34
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
cannam@34
|
2
|
cannam@7
|
3 // GetKeyMode.cpp: implementation of the CGetKeyMode class.
|
cannam@7
|
4 //
|
cannam@7
|
5 //////////////////////////////////////////////////////////////////////
|
cannam@7
|
6
|
cannam@7
|
7 #include "GetKeyMode.h"
|
cannam@16
|
8 #include "maths/MathUtilities.h"
|
cannam@9
|
9 #include "base/Pitch.h"
|
cannam@9
|
10
|
cannam@9
|
11 #include <iostream>
|
cannam@7
|
12
|
cannam@47
|
13 #include <cstring>
|
cannam@47
|
14 #include <cstdlib>
|
cannam@47
|
15
|
cannam@7
|
16 // Chords profile
|
cannam@7
|
17 static double MajProfile[36] =
|
cannam@7
|
18 { 0.0384, 0.0629, 0.0258, 0.0121, 0.0146, 0.0106, 0.0364, 0.0610, 0.0267,
|
chriss@15
|
19 0.0126, 0.0121, 0.0086, 0.0364, 0.0623, 0.0279, 0.0275, 0.0414, 0.0186,
|
cannam@7
|
20 0.0173, 0.0248, 0.0145, 0.0364, 0.0631, 0.0262, 0.0129, 0.0150, 0.0098,
|
cannam@7
|
21 0.0312, 0.0521, 0.0235, 0.0129, 0.0142, 0.0095, 0.0289, 0.0478, 0.0239};
|
cannam@7
|
22
|
cannam@7
|
23 static double MinProfile[36] =
|
cannam@7
|
24 { 0.0375, 0.0682, 0.0299, 0.0119, 0.0138, 0.0093, 0.0296, 0.0543, 0.0257,
|
cannam@7
|
25 0.0292, 0.0519, 0.0246, 0.0159, 0.0234, 0.0135, 0.0291, 0.0544, 0.0248,
|
cannam@7
|
26 0.0137, 0.0176, 0.0104, 0.0352, 0.0670, 0.0302, 0.0222, 0.0349, 0.0164,
|
cannam@7
|
27 0.0174, 0.0297, 0.0166, 0.0222, 0.0401, 0.0202, 0.0175, 0.0270, 0.0146};
|
cannam@7
|
28 //
|
cannam@40
|
29
|
cannam@7
|
30
|
cannam@7
|
31 //////////////////////////////////////////////////////////////////////
|
cannam@7
|
32 // Construction/Destruction
|
cannam@7
|
33 //////////////////////////////////////////////////////////////////////
|
cannam@7
|
34
|
cannam@9
|
35 GetKeyMode::GetKeyMode( int sampleRate, float tuningFrequency,
|
cannam@34
|
36 double hpcpAverage, double medianAverage ) :
|
cannam@34
|
37 m_hpcpAverage( hpcpAverage ),
|
cannam@34
|
38 m_medianAverage( medianAverage ),
|
cannam@34
|
39 m_ChrPointer(0),
|
cannam@34
|
40 m_DecimatedBuffer(0),
|
cannam@34
|
41 m_ChromaBuffer(0),
|
cannam@34
|
42 m_MeanHPCP(0),
|
cannam@34
|
43 m_MajCorr(0),
|
cannam@34
|
44 m_MinCorr(0),
|
cannam@34
|
45 m_Keys(0),
|
cannam@34
|
46 m_MedianFilterBuffer(0),
|
cannam@40
|
47 m_SortedBuffer(0),
|
cannam@40
|
48 m_keyStrengths(0)
|
cannam@7
|
49 {
|
cannam@34
|
50 m_DecimationFactor = 8;
|
cannam@34
|
51
|
cannam@34
|
52 // Chromagram configuration parameters
|
cannam@34
|
53 m_ChromaConfig.normalise = MathUtilities::NormaliseUnitMax;
|
cannam@34
|
54 m_ChromaConfig.FS = lrint(sampleRate/(double)m_DecimationFactor);
|
cannam@58
|
55 if (m_ChromaConfig.FS < 1) m_ChromaConfig.FS = 1;
|
cannam@7
|
56
|
cannam@34
|
57 // Set C (= MIDI #12) as our base :
|
cannam@34
|
58 // This implies that key = 1 => Cmaj, key = 12 => Bmaj, key = 13 => Cmin, etc.
|
cannam@34
|
59 m_ChromaConfig.min = Pitch::getFrequencyForPitch
|
cannam@40
|
60 (48, 0, tuningFrequency);
|
cannam@34
|
61 m_ChromaConfig.max = Pitch::getFrequencyForPitch
|
cannam@34
|
62 (96, 0, tuningFrequency);
|
cannam@9
|
63
|
cannam@34
|
64 m_ChromaConfig.BPO = 36;
|
cannam@34
|
65 m_ChromaConfig.CQThresh = 0.0054;
|
cannam@9
|
66
|
cannam@34
|
67 // Chromagram inst.
|
cannam@34
|
68 m_Chroma = new Chromagram( m_ChromaConfig );
|
cannam@10
|
69
|
cannam@34
|
70 // Get calculated parameters from chroma object
|
cannam@34
|
71 m_ChromaFrameSize = m_Chroma->getFrameSize();
|
cannam@34
|
72 // override hopsize for this application
|
cannam@34
|
73 m_ChromaHopSize = m_ChromaFrameSize;
|
cannam@34
|
74 m_BPO = m_ChromaConfig.BPO;
|
cannam@10
|
75
|
cannam@40
|
76 // std::cerr << "chroma frame size = " << m_ChromaFrameSize << ", decimation factor = " << m_DecimationFactor << " therefore block size = " << getBlockSize() << std::endl;
|
cannam@40
|
77
|
cannam@34
|
78 // Chromagram average and estimated key median filter lengths
|
cannam@34
|
79 m_ChromaBuffersize = (int)ceil( m_hpcpAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
|
cannam@34
|
80 m_MedianWinsize = (int)ceil( m_medianAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
|
cannam@34
|
81
|
cannam@34
|
82 // Reset counters
|
cannam@34
|
83 m_bufferindex = 0;
|
cannam@34
|
84 m_ChromaBufferFilling = 0;
|
cannam@34
|
85 m_MedianBufferFilling = 0;
|
cannam@9
|
86
|
cannam@34
|
87 // Spawn objectc/arrays
|
cannam@34
|
88 m_DecimatedBuffer = new double[m_ChromaFrameSize];
|
cannam@34
|
89
|
cannam@34
|
90 m_ChromaBuffer = new double[m_BPO * m_ChromaBuffersize];
|
cannam@34
|
91 memset( m_ChromaBuffer, 0, sizeof(double) * m_BPO * m_ChromaBuffersize);
|
cannam@34
|
92
|
cannam@34
|
93 m_MeanHPCP = new double[m_BPO];
|
cannam@34
|
94
|
cannam@34
|
95 m_MajCorr = new double[m_BPO];
|
cannam@34
|
96 m_MinCorr = new double[m_BPO];
|
cannam@34
|
97 m_Keys = new double[2*m_BPO];
|
cannam@34
|
98
|
cannam@34
|
99 m_MedianFilterBuffer = new int[ m_MedianWinsize ];
|
cannam@34
|
100 memset( m_MedianFilterBuffer, 0, sizeof(int)*m_MedianWinsize);
|
cannam@34
|
101
|
cannam@34
|
102 m_SortedBuffer = new int[ m_MedianWinsize ];
|
cannam@34
|
103 memset( m_SortedBuffer, 0, sizeof(int)*m_MedianWinsize);
|
cannam@34
|
104
|
cannam@34
|
105 m_Decimator = new Decimator
|
cannam@34
|
106 ( m_ChromaFrameSize*m_DecimationFactor, m_DecimationFactor );
|
cannam@40
|
107
|
cannam@40
|
108 m_keyStrengths = new double[24];
|
cannam@7
|
109 }
|
cannam@7
|
110
|
cannam@7
|
111 GetKeyMode::~GetKeyMode()
|
cannam@7
|
112 {
|
cannam@7
|
113
|
cannam@34
|
114 delete m_Chroma;
|
cannam@34
|
115 delete m_Decimator;
|
cannam@34
|
116
|
cannam@34
|
117 delete [] m_DecimatedBuffer;
|
cannam@34
|
118 delete [] m_ChromaBuffer;
|
cannam@34
|
119 delete [] m_MeanHPCP;
|
cannam@34
|
120 delete [] m_MajCorr;
|
cannam@34
|
121 delete [] m_MinCorr;
|
cannam@34
|
122 delete [] m_Keys;
|
cannam@34
|
123 delete [] m_MedianFilterBuffer;
|
cannam@34
|
124 delete [] m_SortedBuffer;
|
cannam@40
|
125
|
cannam@40
|
126 delete[] m_keyStrengths;
|
cannam@7
|
127 }
|
cannam@7
|
128
|
cannam@7
|
129 double GetKeyMode::krumCorr(double *pData1, double *pData2, unsigned int length)
|
cannam@7
|
130 {
|
cannam@34
|
131 double retVal= 0.0;
|
cannam@34
|
132
|
cannam@34
|
133 double num = 0;
|
cannam@34
|
134 double den = 0;
|
cannam@34
|
135 double mX = MathUtilities::mean( pData1, length );
|
cannam@34
|
136 double mY = MathUtilities::mean( pData2, length );
|
cannam@34
|
137
|
cannam@34
|
138 double sum1 = 0;
|
cannam@34
|
139 double sum2 = 0;
|
cannam@34
|
140
|
cannam@34
|
141 for( unsigned int i = 0; i <length; i++ )
|
cannam@34
|
142 {
|
cannam@34
|
143 num += ( pData1[i] - mX ) * ( pData2[i] - mY );
|
cannam@7
|
144
|
cannam@34
|
145 sum1 += ( (pData1[i]-mX) * (pData1[i]-mX) );
|
cannam@34
|
146 sum2 += ( (pData2[i]-mY) * (pData2[i]-mY) );
|
cannam@34
|
147 }
|
cannam@34
|
148
|
cannam@34
|
149 den = sqrt(sum1 * sum2);
|
cannam@34
|
150
|
cannam@34
|
151 if( den>0 )
|
cannam@34
|
152 retVal = num/den;
|
cannam@34
|
153 else
|
cannam@34
|
154 retVal = 0;
|
cannam@7
|
155
|
cannam@7
|
156
|
cannam@34
|
157 return retVal;
|
cannam@7
|
158 }
|
cannam@7
|
159
|
cannam@7
|
160 int GetKeyMode::process(double *PCMData)
|
cannam@7
|
161 {
|
cannam@34
|
162 int key;
|
cannam@7
|
163
|
cannam@34
|
164 unsigned int j,k;
|
cannam@7
|
165
|
cannam@34
|
166 //////////////////////////////////////////////
|
cannam@34
|
167 m_Decimator->process( PCMData, m_DecimatedBuffer);
|
cannam@7
|
168
|
cannam@34
|
169 m_ChrPointer = m_Chroma->process( m_DecimatedBuffer );
|
cannam@7
|
170
|
chriss@15
|
171
|
cannam@34
|
172 // Move bins such that the centre of the base note is in the
|
cannam@34
|
173 // middle of its three bins :
|
cannam@34
|
174 // Added 21.11.07 by Chris Sutton based on debugging with Katy
|
cannam@34
|
175 // Noland + comparison with Matlab equivalent.
|
cannam@34
|
176 MathUtilities::circShift( m_ChrPointer, m_BPO, 1);
|
cannam@23
|
177 /*
|
cannam@73
|
178 std::cout << "raw chroma: ";
|
cannam@73
|
179 for (int ii = 0; ii < m_BPO; ++ii) {
|
cannam@73
|
180 if (ii % (m_BPO/12) == 0) std::cout << "\n";
|
cannam@73
|
181 std::cout << m_ChrPointer[ii] << " ";
|
cannam@73
|
182 }
|
cannam@73
|
183 std::cout << std::endl;
|
cannam@23
|
184 */
|
cannam@34
|
185 // populate hpcp values;
|
cannam@34
|
186 int cbidx;
|
cannam@34
|
187 for( j = 0; j < m_BPO; j++ )
|
cannam@34
|
188 {
|
cannam@34
|
189 cbidx = (m_bufferindex * m_BPO) + j;
|
cannam@34
|
190 m_ChromaBuffer[ cbidx ] = m_ChrPointer[j];
|
cannam@34
|
191 }
|
cannam@7
|
192
|
cannam@34
|
193 //keep track of input buffers;
|
cannam@34
|
194 if( m_bufferindex++ >= m_ChromaBuffersize - 1)
|
cannam@34
|
195 m_bufferindex = 0;
|
cannam@7
|
196
|
cannam@34
|
197 // track filling of chroma matrix
|
cannam@34
|
198 if( m_ChromaBufferFilling++ >= m_ChromaBuffersize)
|
cannam@34
|
199 m_ChromaBufferFilling = m_ChromaBuffersize;
|
cannam@7
|
200
|
cannam@34
|
201 //calculate mean
|
cannam@34
|
202 for( k = 0; k < m_BPO; k++ )
|
cannam@34
|
203 {
|
cannam@34
|
204 double mnVal = 0.0;
|
cannam@34
|
205 for( j = 0; j < m_ChromaBufferFilling; j++ )
|
cannam@34
|
206 {
|
cannam@34
|
207 mnVal += m_ChromaBuffer[ k + (j*m_BPO) ];
|
cannam@34
|
208 }
|
cannam@7
|
209
|
cannam@34
|
210 m_MeanHPCP[k] = mnVal/(double)m_ChromaBufferFilling;
|
cannam@34
|
211 }
|
cannam@7
|
212
|
cannam@7
|
213
|
cannam@34
|
214 for( k = 0; k < m_BPO; k++ )
|
cannam@34
|
215 {
|
cannam@34
|
216 m_MajCorr[k] = krumCorr( m_MeanHPCP, MajProfile, m_BPO );
|
cannam@34
|
217 m_MinCorr[k] = krumCorr( m_MeanHPCP, MinProfile, m_BPO );
|
cannam@7
|
218
|
cannam@34
|
219 MathUtilities::circShift( MajProfile, m_BPO, 1 );
|
cannam@34
|
220 MathUtilities::circShift( MinProfile, m_BPO, 1 );
|
cannam@34
|
221 }
|
chriss@15
|
222
|
cannam@34
|
223 for( k = 0; k < m_BPO; k++ )
|
cannam@34
|
224 {
|
cannam@34
|
225 m_Keys[k] = m_MajCorr[k];
|
cannam@34
|
226 m_Keys[k+m_BPO] = m_MinCorr[k];
|
cannam@34
|
227 }
|
chriss@15
|
228
|
cannam@40
|
229 for (k = 0; k < 24; ++k) {
|
cannam@40
|
230 m_keyStrengths[k] = 0;
|
cannam@40
|
231 }
|
cannam@40
|
232
|
cannam@40
|
233 for( k = 0; k < m_BPO*2; k++ )
|
cannam@40
|
234 {
|
cannam@73
|
235 int idx = k / (m_BPO/12);
|
cannam@73
|
236 int rem = k % (m_BPO/12);
|
cannam@73
|
237 if (rem == 0 || m_Keys[k] > m_keyStrengths[idx]) {
|
cannam@73
|
238 m_keyStrengths[idx] = m_Keys[k];
|
cannam@73
|
239 }
|
cannam@73
|
240
|
cannam@73
|
241 // m_keyStrengths[k/(m_BPO/12)] += m_Keys[k];
|
cannam@40
|
242 }
|
chriss@15
|
243
|
cannam@23
|
244 /*
|
cannam@34
|
245 std::cout << "raw keys: ";
|
cannam@34
|
246 for (int ii = 0; ii < 2*m_BPO; ++ii) {
|
cannam@73
|
247 if (ii % (m_BPO/12) == 0) std::cout << "\n";
|
cannam@73
|
248 std::cout << m_Keys[ii] << " ";
|
cannam@73
|
249 }
|
cannam@73
|
250 std::cout << std::endl;
|
cannam@73
|
251
|
cannam@73
|
252 std::cout << "key strengths: ";
|
cannam@73
|
253 for (int ii = 0; ii < 24; ++ii) {
|
cannam@73
|
254 if (ii % 6 == 0) std::cout << "\n";
|
cannam@73
|
255 std::cout << m_keyStrengths[ii] << " ";
|
cannam@34
|
256 }
|
cannam@34
|
257 std::cout << std::endl;
|
cannam@23
|
258 */
|
cannam@34
|
259 double dummy;
|
cannam@34
|
260 // '1 +' because we number keys 1-24, not 0-23.
|
cannam@34
|
261 key = 1 + (int)ceil( (double)MathUtilities::getMax( m_Keys, 2* m_BPO, &dummy )/3 );
|
cannam@9
|
262
|
cannam@73
|
263 // std::cout << "key pre-sorting: " << key << std::endl;
|
cannam@7
|
264
|
cannam@7
|
265
|
cannam@34
|
266 //Median filtering
|
cannam@7
|
267
|
cannam@34
|
268 // track Median buffer initial filling
|
cannam@34
|
269 if( m_MedianBufferFilling++ >= m_MedianWinsize)
|
cannam@34
|
270 m_MedianBufferFilling = m_MedianWinsize;
|
cannam@7
|
271
|
cannam@34
|
272 //shift median buffer
|
cannam@34
|
273 for( k = 1; k < m_MedianWinsize; k++ )
|
cannam@34
|
274 {
|
cannam@34
|
275 m_MedianFilterBuffer[ k - 1 ] = m_MedianFilterBuffer[ k ];
|
cannam@34
|
276 }
|
cannam@7
|
277
|
cannam@34
|
278 //write new key value into median buffer
|
cannam@34
|
279 m_MedianFilterBuffer[ m_MedianWinsize - 1 ] = key;
|
cannam@7
|
280
|
cannam@7
|
281
|
cannam@34
|
282 //Copy median into sorting buffer, reversed
|
cannam@34
|
283 unsigned int ijx = 0;
|
cannam@34
|
284 for( k = 0; k < m_MedianWinsize; k++ )
|
cannam@34
|
285 {
|
cannam@34
|
286 m_SortedBuffer[k] = m_MedianFilterBuffer[m_MedianWinsize-1-ijx];
|
cannam@34
|
287 ijx++;
|
cannam@34
|
288 }
|
cannam@7
|
289
|
cannam@34
|
290 qsort(m_SortedBuffer, m_MedianBufferFilling, sizeof(unsigned int),
|
cannam@34
|
291 MathUtilities::compareInt);
|
cannam@34
|
292 /*
|
cannam@34
|
293 std::cout << "sorted: ";
|
cannam@34
|
294 for (int ii = 0; ii < m_MedianBufferFilling; ++ii) {
|
cannam@34
|
295 std::cout << m_SortedBuffer[ii] << " ";
|
cannam@34
|
296 }
|
cannam@34
|
297 std::cout << std::endl;
|
cannam@34
|
298 */
|
cannam@34
|
299 int sortlength = m_MedianBufferFilling;
|
cannam@34
|
300 int midpoint = (int)ceil((double)sortlength/2);
|
cannam@7
|
301
|
cannam@73
|
302 // std::cout << "midpoint = " << midpoint << endl;
|
cannam@73
|
303
|
cannam@34
|
304 if( midpoint <= 0 )
|
cannam@34
|
305 midpoint = 1;
|
cannam@7
|
306
|
cannam@34
|
307 key = m_SortedBuffer[midpoint-1];
|
cannam@7
|
308
|
cannam@73
|
309 // std::cout << "returning key = " << key << endl;
|
cannam@73
|
310
|
cannam@34
|
311 return key;
|
cannam@7
|
312 }
|
cannam@7
|
313
|
cannam@7
|
314
|
cannam@43
|
315 bool GetKeyMode::isModeMinor( int key )
|
cannam@7
|
316 {
|
cannam@34
|
317 return (key > 12);
|
cannam@7
|
318 }
|