annotate dsp/keydetection/GetKeyMode.cpp @ 236:ac95b396c8c3

...
author Chris Cannam <c.cannam@qmul.ac.uk>
date Mon, 02 Apr 2007 13:20:30 +0000
parents 83ccb61f9c0c
children 10c3f9df4a07
rev   line source
c@232 1 // GetKeyMode.cpp: implementation of the CGetKeyMode class.
c@232 2 //
c@232 3 //////////////////////////////////////////////////////////////////////
c@232 4
c@232 5 #include "GetKeyMode.h"
c@232 6 #include "dsp/maths/MathUtilities.h"
c@234 7 #include "base/Pitch.h"
c@234 8
c@234 9 #include <iostream>
c@232 10
c@232 11 // Chords profile
c@232 12 static double MajProfile[36] =
c@232 13 { 0.0384, 0.0629, 0.0258, 0.0121, 0.0146, 0.0106, 0.0364, 0.0610, 0.0267,
c@232 14 0.0126, 0.0121, 0.0086, 0.0364, 0.0623, 0.0279, 0.0275, 0.0414, 0.0186,
c@232 15 0.0173, 0.0248, 0.0145, 0.0364, 0.0631, 0.0262, 0.0129, 0.0150, 0.0098,
c@232 16 0.0312, 0.0521, 0.0235, 0.0129, 0.0142, 0.0095, 0.0289, 0.0478, 0.0239};
c@232 17
c@232 18 static double MinProfile[36] =
c@232 19 { 0.0375, 0.0682, 0.0299, 0.0119, 0.0138, 0.0093, 0.0296, 0.0543, 0.0257,
c@232 20 0.0292, 0.0519, 0.0246, 0.0159, 0.0234, 0.0135, 0.0291, 0.0544, 0.0248,
c@232 21 0.0137, 0.0176, 0.0104, 0.0352, 0.0670, 0.0302, 0.0222, 0.0349, 0.0164,
c@232 22 0.0174, 0.0297, 0.0166, 0.0222, 0.0401, 0.0202, 0.0175, 0.0270, 0.0146};
c@232 23 //
c@232 24
c@232 25 //////////////////////////////////////////////////////////////////////
c@232 26 // Construction/Destruction
c@232 27 //////////////////////////////////////////////////////////////////////
c@232 28
c@234 29 GetKeyMode::GetKeyMode( int sampleRate, float tuningFrequency,
c@234 30 double hpcpAverage, double medianAverage )
c@232 31 :
c@232 32 m_hpcpAverage( hpcpAverage ),
c@232 33 m_medianAverage( medianAverage ),
c@232 34 m_ChrPointer(0),
c@232 35 m_DecimatedBuffer(0),
c@232 36 m_ChromaBuffer(0),
c@232 37 m_MeanHPCP(0),
c@232 38 m_MajCorr(0),
c@232 39 m_MinCorr(0),
c@232 40 m_Keys(0),
c@232 41 m_MedianFilterBuffer(0),
c@232 42 m_SortedBuffer(0)
c@232 43 {
c@232 44 m_DecimationFactor = 8;
c@232 45
c@232 46 //Chromagram configuration parameters
c@234 47 m_ChromaConfig.isNormalised = 1;
c@234 48 m_ChromaConfig.FS = lrint(sampleRate/(double)m_DecimationFactor);
c@234 49
c@234 50 // m_ChromaConfig.min = 111.0641;
c@234 51 // m_ChromaConfig.max = 1.7770e+003;
c@234 52
c@235 53 // m_ChromaConfig.min = Pitch::getFrequencyForPitch
c@235 54 // (12, 0, tuningFrequency);
c@235 55 // m_ChromaConfig.max = Pitch::getFrequencyForPitch
c@235 56 // (96, 0, tuningFrequency);
c@235 57
c@235 58 // The chromagram minimum pitch is 1/6 of a tone above A, two
c@235 59 // octaves below middle C (for a 36-bin chromagram). The
c@235 60 // maximum pitch is four octaves higher.
c@235 61
c@234 62 m_ChromaConfig.min = Pitch::getFrequencyForPitch
c@235 63 (45, 1.f / 3.f, tuningFrequency);
c@235 64
c@235 65 m_ChromaConfig.max = m_ChromaConfig.min * 2;
c@235 66 m_ChromaConfig.max = m_ChromaConfig.max * 2;
c@235 67 m_ChromaConfig.max = m_ChromaConfig.max * 2;
c@235 68 m_ChromaConfig.max = m_ChromaConfig.max * 2;
c@235 69
c@235 70 std::cerr << "Chromagram range: " << m_ChromaConfig.min << " -> " << m_ChromaConfig.max << std::endl;
c@234 71
c@234 72 m_ChromaConfig.BPO = 36;
c@234 73 m_ChromaConfig.CQThresh = 0.0054;
c@232 74
c@232 75 //Chromagram inst.
c@234 76 m_Chroma = new Chromagram( m_ChromaConfig );
c@232 77
c@232 78 //Get calculated parameters from chroma object
c@232 79 m_ChromaFrameSize = m_Chroma->getFrameSize();
c@232 80 //override hopsize for this application
c@232 81 m_ChromaHopSize = m_ChromaFrameSize;//m_Chroma->GetHopSize();
c@234 82 m_BPO = m_ChromaConfig.BPO;
c@232 83
c@232 84 //Chromagram average and estimated key median filter lengths
c@234 85 m_ChromaBuffersize = (int)ceil( m_hpcpAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
c@234 86 m_MedianWinsize = (int)ceil( m_medianAverage * m_ChromaConfig.FS/m_ChromaFrameSize );
c@232 87
c@232 88 //Reset counters
c@232 89 m_bufferindex = 0;
c@232 90 m_ChromaBufferFilling = 0;
c@232 91 m_MedianBufferFilling = 0;
c@232 92
c@232 93
c@232 94 //Spawn objectc/arrays
c@232 95 m_DecimatedBuffer = new double[m_ChromaFrameSize];
c@232 96
c@232 97 m_ChromaBuffer = new double[m_BPO * m_ChromaBuffersize];
c@232 98 memset( m_ChromaBuffer, 0, sizeof(double) * m_BPO * m_ChromaBuffersize);
c@232 99
c@232 100 m_MeanHPCP = new double[m_BPO];
c@232 101
c@232 102 m_MajCorr = new double[m_BPO];
c@232 103 m_MinCorr = new double[m_BPO];
c@232 104 m_Keys = new double[2*m_BPO];
c@232 105
c@232 106 m_MedianFilterBuffer = new int[ m_MedianWinsize ];
c@232 107 memset( m_MedianFilterBuffer, 0, sizeof(int)*m_MedianWinsize);
c@232 108
c@232 109 m_SortedBuffer = new int[ m_MedianWinsize ];
c@232 110 memset( m_SortedBuffer, 0, sizeof(int)*m_MedianWinsize);
c@232 111
c@232 112 m_Decimator = new Decimator( m_ChromaFrameSize*m_DecimationFactor, m_DecimationFactor );
c@232 113 }
c@232 114
c@232 115 GetKeyMode::~GetKeyMode()
c@232 116 {
c@232 117
c@232 118 delete m_Chroma;
c@232 119 delete m_Decimator;
c@232 120
c@232 121 delete [] m_DecimatedBuffer;
c@232 122 delete [] m_ChromaBuffer;
c@232 123 delete [] m_MeanHPCP;
c@232 124 delete [] m_MajCorr;
c@232 125 delete [] m_MinCorr;
c@232 126 delete [] m_Keys;
c@232 127 delete [] m_MedianFilterBuffer;
c@232 128 delete [] m_SortedBuffer;
c@232 129 }
c@232 130
c@232 131 double GetKeyMode::krumCorr(double *pData1, double *pData2, unsigned int length)
c@232 132 {
c@232 133 double retVal= 0.0;
c@232 134
c@232 135 double num = 0;
c@232 136 double den = 0;
c@232 137 double mX = MathUtilities::mean( pData1, length );
c@232 138 double mY = MathUtilities::mean( pData2, length );
c@232 139
c@232 140 double sum1 = 0;
c@232 141 double sum2 = 0;
c@232 142
c@232 143 for( unsigned int i = 0; i <length; i++ )
c@232 144 {
c@232 145 num += ( pData1[i] - mX ) * ( pData2[i] - mY );
c@232 146
c@232 147 sum1 += ( (pData1[i]-mX) * (pData1[i]-mX) );
c@232 148 sum2 += ( (pData2[i]-mY) * (pData2[i]-mY) );
c@232 149 }
c@232 150
c@232 151 den = sqrt(sum1 * sum2);
c@232 152
c@232 153 if( den>0 )
c@232 154 retVal = num/den;
c@232 155 else
c@232 156 retVal = 0;
c@232 157
c@232 158
c@232 159 return retVal;
c@232 160 }
c@232 161
c@232 162 int GetKeyMode::process(double *PCMData)
c@232 163 {
c@232 164 int key;
c@232 165
c@232 166 unsigned int j,k;
c@232 167
c@232 168 //////////////////////////////////////////////
c@232 169 m_Decimator->process( PCMData, m_DecimatedBuffer);
c@232 170
c@232 171 m_ChrPointer = m_Chroma->process( m_DecimatedBuffer );
c@232 172
c@234 173 std::cout << "raw chroma: ";
c@234 174 for (int ii = 0; ii < m_BPO; ++ii) {
c@234 175 std::cout << m_ChrPointer[ii] << " ";
c@234 176 }
c@234 177 std::cout << std::endl;
c@234 178
c@232 179 // populate hpcp values;
c@232 180 int cbidx;
c@232 181 for( j = 0; j < m_BPO; j++ )
c@232 182 {
c@232 183 cbidx = (m_bufferindex * m_BPO) + j;
c@232 184 m_ChromaBuffer[ cbidx ] = m_ChrPointer[j];
c@232 185 }
c@232 186
c@232 187 //keep track of input buffers;
c@232 188 if( m_bufferindex++ >= m_ChromaBuffersize - 1)
c@232 189 m_bufferindex = 0;
c@232 190
c@232 191 // track filling of chroma matrix
c@232 192 if( m_ChromaBufferFilling++ >= m_ChromaBuffersize)
c@232 193 m_ChromaBufferFilling = m_ChromaBuffersize;
c@232 194
c@232 195 //calculate mean
c@232 196 for( k = 0; k < m_BPO; k++ )
c@232 197 {
c@232 198 double mnVal = 0.0;
c@232 199 for( j = 0; j < m_ChromaBufferFilling; j++ )
c@232 200 {
c@232 201 mnVal += m_ChromaBuffer[ k + (j*m_BPO) ];
c@232 202 }
c@232 203
c@232 204 m_MeanHPCP[k] = mnVal/(double)m_ChromaBufferFilling;
c@232 205 }
c@232 206
c@232 207
c@232 208 for( k = 0; k < m_BPO; k++ )
c@232 209 {
c@232 210 m_MajCorr[k] = krumCorr( m_MeanHPCP, MajProfile, m_BPO );
c@232 211 m_MinCorr[k] = krumCorr( m_MeanHPCP, MinProfile, m_BPO );
c@232 212
c@232 213 MathUtilities::circShift( MajProfile, m_BPO, 1 );
c@232 214 MathUtilities::circShift( MinProfile, m_BPO, 1 );
c@232 215 }
c@232 216
c@232 217 for( k = 0; k < m_BPO; k++ )
c@232 218 {
c@232 219 m_Keys[k] = m_MajCorr[k];
c@232 220 m_Keys[k+m_BPO] = m_MinCorr[k];
c@232 221 }
c@236 222 /*
c@234 223 std::cout << "raw keys: ";
c@234 224 for (int ii = 0; ii < 2*m_BPO; ++ii) {
c@234 225 std::cout << m_Keys[ii] << " ";
c@234 226 }
c@234 227 std::cout << std::endl;
c@236 228 */
c@232 229 double dummy;
c@234 230 key = /*1 +*/ (int)ceil( (double)MathUtilities::getMax( m_Keys, 2* m_BPO, &dummy )/3 );
c@234 231
c@236 232 // std::cout << "key pre-sorting: " << key << std::endl;
c@232 233
c@232 234
c@232 235 //Median filtering
c@232 236
c@232 237 // track Median buffer initial filling
c@232 238 if( m_MedianBufferFilling++ >= m_MedianWinsize)
c@232 239 m_MedianBufferFilling = m_MedianWinsize;
c@232 240
c@232 241 //shift median buffer
c@232 242 for( k = 1; k < m_MedianWinsize; k++ )
c@232 243 {
c@232 244 m_MedianFilterBuffer[ k - 1 ] = m_MedianFilterBuffer[ k ];
c@232 245 }
c@232 246
c@232 247 //write new key value into median buffer
c@232 248 m_MedianFilterBuffer[ m_MedianWinsize - 1 ] = key;
c@232 249
c@232 250
c@232 251 //Copy median into sorting buffer, reversed
c@232 252 unsigned int ijx = 0;
c@232 253 for( k = 0; k < m_MedianWinsize; k++ )
c@232 254 {
c@232 255 m_SortedBuffer[k] = m_MedianFilterBuffer[m_MedianWinsize-1-ijx];
c@232 256 ijx++;
c@232 257 }
c@232 258
c@232 259
c@232 260 //quicksort
c@232 261 qsort(m_SortedBuffer, m_MedianBufferFilling, sizeof(unsigned int), MathUtilities::compareInt);
c@236 262 /*
c@234 263 std::cout << "sorted: ";
c@234 264 for (int ii = 0; ii < m_MedianBufferFilling; ++ii) {
c@234 265 std::cout << m_SortedBuffer[ii] << " ";
c@234 266 }
c@234 267 std::cout << std::endl;
c@236 268 */
c@232 269 int sortlength = m_MedianBufferFilling;
c@232 270 int midpoint = (int)ceil((double)sortlength/2);
c@232 271
c@232 272 if( midpoint <= 0 )
c@232 273 midpoint = 1;
c@232 274
c@232 275 key = m_SortedBuffer[midpoint-1];
c@232 276
c@232 277 return key;
c@232 278 }
c@232 279
c@232 280
c@232 281 int GetKeyMode::isModeMinor( int key )
c@232 282 {
c@233 283 // return ((key-1 - (int)MathUtilities::mod((double)(key-1),(double)12))/12);
c@233 284 return (key > 12);
c@232 285 }