c@225
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
c@225
|
2
|
c@225
|
3 /*
|
c@225
|
4 QM DSP Library
|
c@225
|
5
|
c@225
|
6 Centre for Digital Music, Queen Mary, University of London.
|
c@309
|
7 This file 2005-2006 Christian Landone.
|
c@309
|
8
|
c@309
|
9 This program is free software; you can redistribute it and/or
|
c@309
|
10 modify it under the terms of the GNU General Public License as
|
c@309
|
11 published by the Free Software Foundation; either version 2 of the
|
c@309
|
12 License, or (at your option) any later version. See the file
|
c@309
|
13 COPYING included with this distribution for more information.
|
c@225
|
14 */
|
c@225
|
15
|
c@225
|
16 #include "DetectionFunction.h"
|
c@272
|
17 #include <cstring>
|
c@225
|
18
|
c@225
|
19 //////////////////////////////////////////////////////////////////////
|
c@225
|
20 // Construction/Destruction
|
c@225
|
21 //////////////////////////////////////////////////////////////////////
|
c@225
|
22
|
c@225
|
23 DetectionFunction::DetectionFunction( DFConfig Config ) :
|
c@225
|
24 m_window(0)
|
c@225
|
25 {
|
c@227
|
26 m_magHistory = NULL;
|
c@227
|
27 m_phaseHistory = NULL;
|
c@227
|
28 m_phaseHistoryOld = NULL;
|
c@239
|
29 m_magPeaks = NULL;
|
c@225
|
30
|
c@225
|
31 initialise( Config );
|
c@225
|
32 }
|
c@225
|
33
|
c@225
|
34 DetectionFunction::~DetectionFunction()
|
c@225
|
35 {
|
c@225
|
36 deInitialise();
|
c@225
|
37 }
|
c@225
|
38
|
c@225
|
39
|
c@225
|
40 void DetectionFunction::initialise( DFConfig Config )
|
c@225
|
41 {
|
c@225
|
42 m_dataLength = Config.frameLength;
|
c@340
|
43 m_halfLength = m_dataLength/2 + 1;
|
c@239
|
44
|
c@225
|
45 m_DFType = Config.DFType;
|
c@238
|
46 m_stepSize = Config.stepSize;
|
c@410
|
47 m_dbRise = Config.dbRise;
|
c@225
|
48
|
c@239
|
49 m_whiten = Config.adaptiveWhitening;
|
c@239
|
50 m_whitenRelaxCoeff = Config.whiteningRelaxCoeff;
|
c@239
|
51 m_whitenFloor = Config.whiteningFloor;
|
c@239
|
52 if (m_whitenRelaxCoeff < 0) m_whitenRelaxCoeff = 0.9997;
|
c@239
|
53 if (m_whitenFloor < 0) m_whitenFloor = 0.01;
|
c@239
|
54
|
c@227
|
55 m_magHistory = new double[ m_halfLength ];
|
c@227
|
56 memset(m_magHistory,0, m_halfLength*sizeof(double));
|
c@225
|
57
|
c@227
|
58 m_phaseHistory = new double[ m_halfLength ];
|
c@227
|
59 memset(m_phaseHistory,0, m_halfLength*sizeof(double));
|
c@225
|
60
|
c@227
|
61 m_phaseHistoryOld = new double[ m_halfLength ];
|
c@227
|
62 memset(m_phaseHistoryOld,0, m_halfLength*sizeof(double));
|
c@225
|
63
|
c@239
|
64 m_magPeaks = new double[ m_halfLength ];
|
c@239
|
65 memset(m_magPeaks,0, m_halfLength*sizeof(double));
|
c@239
|
66
|
c@356
|
67 m_phaseVoc = new PhaseVocoder(m_dataLength, m_stepSize);
|
c@225
|
68
|
c@225
|
69 m_magnitude = new double[ m_halfLength ];
|
c@225
|
70 m_thetaAngle = new double[ m_halfLength ];
|
c@340
|
71 m_unwrapped = new double[ m_halfLength ];
|
c@225
|
72
|
c@225
|
73 m_window = new Window<double>(HanningWindow, m_dataLength);
|
c@344
|
74 m_windowed = new double[ m_dataLength ];
|
c@225
|
75 }
|
c@225
|
76
|
c@225
|
77 void DetectionFunction::deInitialise()
|
c@225
|
78 {
|
c@227
|
79 delete [] m_magHistory ;
|
c@227
|
80 delete [] m_phaseHistory ;
|
c@227
|
81 delete [] m_phaseHistoryOld ;
|
c@239
|
82 delete [] m_magPeaks ;
|
c@225
|
83
|
c@225
|
84 delete m_phaseVoc;
|
c@225
|
85
|
c@225
|
86 delete [] m_magnitude;
|
c@225
|
87 delete [] m_thetaAngle;
|
c@344
|
88 delete [] m_windowed;
|
c@344
|
89 delete [] m_unwrapped;
|
c@225
|
90
|
c@225
|
91 delete m_window;
|
c@225
|
92 }
|
c@225
|
93
|
c@344
|
94 double DetectionFunction::processTimeDomain(const double *samples)
|
c@225
|
95 {
|
c@344
|
96 m_window->cut(samples, m_windowed);
|
c@280
|
97
|
c@344
|
98 m_phaseVoc->processTimeDomain(m_windowed,
|
c@344
|
99 m_magnitude, m_thetaAngle, m_unwrapped);
|
c@225
|
100
|
c@239
|
101 if (m_whiten) whiten();
|
c@239
|
102
|
c@227
|
103 return runDF();
|
c@227
|
104 }
|
c@227
|
105
|
c@344
|
106 double DetectionFunction::processFrequencyDomain(const double *reals,
|
c@344
|
107 const double *imags)
|
c@227
|
108 {
|
c@344
|
109 m_phaseVoc->processFrequencyDomain(reals, imags,
|
c@344
|
110 m_magnitude, m_thetaAngle, m_unwrapped);
|
c@227
|
111
|
c@239
|
112 if (m_whiten) whiten();
|
c@239
|
113
|
c@227
|
114 return runDF();
|
c@227
|
115 }
|
c@227
|
116
|
c@239
|
117 void DetectionFunction::whiten()
|
c@239
|
118 {
|
c@239
|
119 for (unsigned int i = 0; i < m_halfLength; ++i) {
|
c@239
|
120 double m = m_magnitude[i];
|
c@239
|
121 if (m < m_magPeaks[i]) {
|
c@239
|
122 m = m + (m_magPeaks[i] - m) * m_whitenRelaxCoeff;
|
c@239
|
123 }
|
c@239
|
124 if (m < m_whitenFloor) m = m_whitenFloor;
|
c@239
|
125 m_magPeaks[i] = m;
|
c@239
|
126 m_magnitude[i] /= m;
|
c@239
|
127 }
|
c@239
|
128 }
|
c@239
|
129
|
c@227
|
130 double DetectionFunction::runDF()
|
c@227
|
131 {
|
c@227
|
132 double retVal = 0;
|
c@227
|
133
|
c@225
|
134 switch( m_DFType )
|
c@225
|
135 {
|
c@225
|
136 case DF_HFC:
|
c@225
|
137 retVal = HFC( m_halfLength, m_magnitude);
|
c@225
|
138 break;
|
c@225
|
139
|
c@238
|
140 case DF_SPECDIFF:
|
c@225
|
141 retVal = specDiff( m_halfLength, m_magnitude);
|
c@225
|
142 break;
|
c@225
|
143
|
c@225
|
144 case DF_PHASEDEV:
|
c@345
|
145 // Using the instantaneous phases here actually provides the
|
c@345
|
146 // same results (for these calculations) as if we had used
|
c@345
|
147 // unwrapped phases, but without the possible accumulation of
|
c@345
|
148 // phase error over time
|
c@239
|
149 retVal = phaseDev( m_halfLength, m_thetaAngle);
|
c@225
|
150 break;
|
c@225
|
151
|
c@225
|
152 case DF_COMPLEXSD:
|
c@225
|
153 retVal = complexSD( m_halfLength, m_magnitude, m_thetaAngle);
|
c@225
|
154 break;
|
c@237
|
155
|
c@237
|
156 case DF_BROADBAND:
|
c@239
|
157 retVal = broadband( m_halfLength, m_magnitude);
|
c@239
|
158 break;
|
c@225
|
159 }
|
c@225
|
160
|
c@225
|
161 return retVal;
|
c@225
|
162 }
|
c@225
|
163
|
c@225
|
164 double DetectionFunction::HFC(unsigned int length, double *src)
|
c@225
|
165 {
|
c@225
|
166 unsigned int i;
|
c@225
|
167 double val = 0;
|
c@225
|
168
|
c@225
|
169 for( i = 0; i < length; i++)
|
c@225
|
170 {
|
c@225
|
171 val += src[ i ] * ( i + 1);
|
c@225
|
172 }
|
c@225
|
173 return val;
|
c@225
|
174 }
|
c@225
|
175
|
c@225
|
176 double DetectionFunction::specDiff(unsigned int length, double *src)
|
c@225
|
177 {
|
c@225
|
178 unsigned int i;
|
c@225
|
179 double val = 0.0;
|
c@225
|
180 double temp = 0.0;
|
c@225
|
181 double diff = 0.0;
|
c@225
|
182
|
c@225
|
183 for( i = 0; i < length; i++)
|
c@225
|
184 {
|
c@227
|
185 temp = fabs( (src[ i ] * src[ i ]) - (m_magHistory[ i ] * m_magHistory[ i ]) );
|
c@225
|
186
|
c@225
|
187 diff= sqrt(temp);
|
c@225
|
188
|
c@238
|
189 // (See note in phaseDev below.)
|
c@238
|
190
|
c@238
|
191 val += diff;
|
c@225
|
192
|
c@227
|
193 m_magHistory[ i ] = src[ i ];
|
c@225
|
194 }
|
c@225
|
195
|
c@225
|
196 return val;
|
c@225
|
197 }
|
c@225
|
198
|
c@225
|
199
|
c@239
|
200 double DetectionFunction::phaseDev(unsigned int length, double *srcPhase)
|
c@225
|
201 {
|
c@225
|
202 unsigned int i;
|
c@225
|
203 double tmpPhase = 0;
|
c@225
|
204 double tmpVal = 0;
|
c@225
|
205 double val = 0;
|
c@225
|
206
|
c@225
|
207 double dev = 0;
|
c@225
|
208
|
c@225
|
209 for( i = 0; i < length; i++)
|
c@225
|
210 {
|
c@227
|
211 tmpPhase = (srcPhase[ i ]- 2*m_phaseHistory[ i ]+m_phaseHistoryOld[ i ]);
|
c@225
|
212 dev = MathUtilities::princarg( tmpPhase );
|
c@238
|
213
|
c@238
|
214 // A previous version of this code only counted the value here
|
c@238
|
215 // if the magnitude exceeded 0.1. My impression is that
|
c@238
|
216 // doesn't greatly improve the results for "loud" music (so
|
c@238
|
217 // long as the peak picker is reasonably sophisticated), but
|
c@238
|
218 // does significantly damage its ability to work with quieter
|
c@238
|
219 // music, so I'm removing it and counting the result always.
|
c@238
|
220 // Same goes for the spectral difference measure above.
|
c@225
|
221
|
c@238
|
222 tmpVal = fabs(dev);
|
c@238
|
223 val += tmpVal ;
|
c@225
|
224
|
c@227
|
225 m_phaseHistoryOld[ i ] = m_phaseHistory[ i ] ;
|
c@227
|
226 m_phaseHistory[ i ] = srcPhase[ i ];
|
c@225
|
227 }
|
c@225
|
228
|
c@225
|
229 return val;
|
c@225
|
230 }
|
c@225
|
231
|
c@225
|
232
|
c@225
|
233 double DetectionFunction::complexSD(unsigned int length, double *srcMagnitude, double *srcPhase)
|
c@225
|
234 {
|
c@225
|
235 unsigned int i;
|
c@225
|
236 double val = 0;
|
c@225
|
237 double tmpPhase = 0;
|
c@225
|
238 double tmpReal = 0;
|
c@225
|
239 double tmpImag = 0;
|
c@225
|
240
|
c@225
|
241 double dev = 0;
|
c@225
|
242 ComplexData meas = ComplexData( 0, 0 );
|
c@227
|
243 ComplexData j = ComplexData( 0, 1 );
|
c@225
|
244
|
c@225
|
245 for( i = 0; i < length; i++)
|
c@225
|
246 {
|
c@227
|
247 tmpPhase = (srcPhase[ i ]- 2*m_phaseHistory[ i ]+m_phaseHistoryOld[ i ]);
|
c@225
|
248 dev= MathUtilities::princarg( tmpPhase );
|
c@225
|
249
|
c@227
|
250 meas = m_magHistory[i] - ( srcMagnitude[ i ] * exp( j * dev) );
|
c@225
|
251
|
c@225
|
252 tmpReal = real( meas );
|
c@225
|
253 tmpImag = imag( meas );
|
c@225
|
254
|
c@225
|
255 val += sqrt( (tmpReal * tmpReal) + (tmpImag * tmpImag) );
|
c@225
|
256
|
c@227
|
257 m_phaseHistoryOld[ i ] = m_phaseHistory[ i ] ;
|
c@227
|
258 m_phaseHistory[ i ] = srcPhase[ i ];
|
c@227
|
259 m_magHistory[ i ] = srcMagnitude[ i ];
|
c@225
|
260 }
|
c@225
|
261
|
c@225
|
262 return val;
|
c@225
|
263 }
|
c@225
|
264
|
c@239
|
265 double DetectionFunction::broadband(unsigned int length, double *src)
|
c@237
|
266 {
|
c@237
|
267 double val = 0;
|
c@237
|
268 for (unsigned int i = 0; i < length; ++i) {
|
c@239
|
269 double sqrmag = src[i] * src[i];
|
c@237
|
270 if (m_magHistory[i] > 0.0) {
|
c@237
|
271 double diff = 10.0 * log10(sqrmag / m_magHistory[i]);
|
c@237
|
272 if (diff > m_dbRise) val = val + 1;
|
c@237
|
273 }
|
c@237
|
274 m_magHistory[i] = sqrmag;
|
c@237
|
275 }
|
c@237
|
276 return val;
|
c@237
|
277 }
|
c@237
|
278
|
c@225
|
279 double* DetectionFunction::getSpectrumMagnitude()
|
c@225
|
280 {
|
c@225
|
281 return m_magnitude;
|
c@225
|
282 }
|
c@225
|
283
|