cannam@0
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
cannam@0
|
2
|
cannam@0
|
3 /*
|
cannam@0
|
4 Vamp feature extraction plugin using the MATCH audio alignment
|
cannam@0
|
5 algorithm.
|
cannam@0
|
6
|
cannam@0
|
7 Centre for Digital Music, Queen Mary, University of London.
|
cannam@0
|
8 This file copyright 2007 Simon Dixon, Chris Cannam and QMUL.
|
cannam@0
|
9
|
cannam@0
|
10 This program is free software; you can redistribute it and/or
|
cannam@0
|
11 modify it under the terms of the GNU General Public License as
|
cannam@0
|
12 published by the Free Software Foundation; either version 2 of the
|
cannam@0
|
13 License, or (at your option) any later version. See the file
|
cannam@0
|
14 COPYING included with this distribution for more information.
|
cannam@0
|
15 */
|
cannam@0
|
16
|
cannam@0
|
17 #include "Matcher.h"
|
cannam@0
|
18 #include "Finder.h"
|
cannam@0
|
19
|
cannam@0
|
20 #include <iostream>
|
cannam@0
|
21
|
cannam@4
|
22 #include <cstdlib>
|
cannam@4
|
23
|
cannam@0
|
24 bool Matcher::silent = true;
|
cannam@0
|
25
|
cannam@7
|
26 const double Matcher::decay = 0.99;
|
cannam@7
|
27 const double Matcher::silenceThreshold = 0.0004;
|
cannam@7
|
28 const int Matcher::MAX_RUN_COUNT = 3;
|
cannam@7
|
29
|
Chris@10
|
30 //#define DEBUG_MATCHER 1
|
Chris@10
|
31
|
cannam@0
|
32 Matcher::Matcher(float rate, Matcher *p)
|
cannam@0
|
33 {
|
Chris@10
|
34 #ifdef DEBUG_MATCHER
|
cannam@0
|
35 std::cerr << "Matcher::Matcher(" << rate << ", " << p << ")" << std::endl;
|
Chris@10
|
36 #endif
|
cannam@0
|
37
|
cannam@0
|
38 sampleRate = rate;
|
cannam@0
|
39 otherMatcher = p; // the first matcher will need this to be set later
|
cannam@0
|
40 firstPM = (!p);
|
cannam@0
|
41 matchFileOffset = 0;
|
cannam@0
|
42 ltAverage = 0;
|
cannam@0
|
43 frameCount = 0;
|
cannam@0
|
44 runCount = 0;
|
cannam@0
|
45 paused = false;
|
cannam@0
|
46 hopSize = 0;
|
cannam@0
|
47 fftSize = 0;
|
cannam@0
|
48 blockSize = 0;
|
cannam@0
|
49 hopTime = 0.020; // DEFAULT, overridden with -h //!!!
|
cannam@0
|
50 fftTime = 0.04644; // DEFAULT, overridden with -f
|
cannam@0
|
51 blockTime = 10.0; // DEFAULT, overridden with -c
|
cannam@0
|
52 normalise1 = true;
|
cannam@0
|
53 normalise2 = false;
|
cannam@0
|
54 normalise3 = false;
|
cannam@0
|
55 normalise4 = true;
|
cannam@0
|
56 useSpectralDifference = true;
|
cannam@0
|
57 useChromaFrequencyMap = false;
|
cannam@0
|
58 scale = 90;
|
cannam@0
|
59 maxFrames = 0; // stop at EOF
|
cannam@0
|
60
|
cannam@0
|
61 hopSize = lrint(sampleRate * hopTime);
|
cannam@7
|
62 fftSize = lrint(pow(2.0, (int)lrint(log(fftTime * sampleRate) / log(2.0))));
|
cannam@0
|
63 blockSize = lrint(blockTime / hopTime);
|
cannam@0
|
64
|
cannam@0
|
65 distance = 0;
|
cannam@0
|
66 bestPathCost = 0;
|
cannam@0
|
67 distYSizes = 0;
|
cannam@0
|
68 distXSize = 0;
|
cannam@0
|
69
|
cannam@0
|
70 initialised = false;
|
cannam@0
|
71
|
cannam@0
|
72 } // default constructor
|
cannam@0
|
73
|
cannam@1
|
74 void
|
cannam@1
|
75 Matcher::setHopSize(int sz)
|
cannam@1
|
76 {
|
cannam@1
|
77 if (initialised) {
|
cannam@1
|
78 std::cerr << "Matcher::setHopSize: Can't set after use" << std::endl;
|
cannam@1
|
79 return;
|
cannam@1
|
80 }
|
cannam@1
|
81
|
cannam@1
|
82 hopSize = sz;
|
cannam@1
|
83 hopTime = float(hopSize) / sampleRate;
|
cannam@1
|
84 blockTime = blockSize * hopTime;
|
cannam@1
|
85 }
|
cannam@1
|
86
|
cannam@0
|
87 Matcher::~Matcher()
|
cannam@0
|
88 {
|
Chris@10
|
89 #ifdef DEBUG_MATCHER
|
cannam@0
|
90 std::cerr << "Matcher(" << this << ")::~Matcher()" << std::endl;
|
Chris@10
|
91 #endif
|
cannam@0
|
92
|
cannam@0
|
93 if (initialised) {
|
cannam@0
|
94
|
cannam@0
|
95 for (int i = 0; i < distXSize; ++i) {
|
cannam@0
|
96 if (distance[i]) {
|
cannam@0
|
97 free(distance[i]);
|
cannam@0
|
98 free(bestPathCost[i]);
|
cannam@0
|
99 }
|
cannam@0
|
100 }
|
cannam@0
|
101 free(distance);
|
cannam@0
|
102 free(bestPathCost);
|
cannam@0
|
103
|
cannam@0
|
104 free(first);
|
cannam@0
|
105 free(last);
|
cannam@0
|
106
|
cannam@0
|
107 free(distYSizes);
|
cannam@0
|
108 }
|
cannam@0
|
109 }
|
cannam@0
|
110
|
cannam@0
|
111 void
|
cannam@0
|
112 Matcher::print()
|
cannam@0
|
113 {
|
cannam@0
|
114 cerr << toString() << endl;
|
cannam@0
|
115 } // print()
|
cannam@0
|
116
|
cannam@0
|
117 string
|
cannam@0
|
118 Matcher::toString()
|
cannam@0
|
119 {
|
cannam@0
|
120 std::stringstream os;
|
cannam@0
|
121 os << "Matcher " << this << ": (" << sampleRate
|
cannam@0
|
122 << "kHz)"
|
cannam@0
|
123 << "\n\tHop size: " << hopSize
|
cannam@0
|
124 << "\n\tFFT size: " << fftSize
|
cannam@0
|
125 << "\n\tBlock size: " << blockSize;
|
cannam@0
|
126 return os.str();
|
cannam@0
|
127 } // toString()
|
cannam@0
|
128
|
cannam@0
|
129 void
|
cannam@0
|
130 Matcher::init()
|
cannam@0
|
131 {
|
cannam@0
|
132 if (initialised) return;
|
cannam@0
|
133
|
cannam@0
|
134 initialised = true;
|
cannam@0
|
135
|
cannam@0
|
136 makeFreqMap(fftSize, sampleRate);
|
cannam@0
|
137
|
cannam@0
|
138 initVector<double>(prevFrame, freqMapSize);
|
cannam@0
|
139 initVector<double>(newFrame, freqMapSize);
|
cannam@0
|
140 initMatrix<double>(frames, blockSize, freqMapSize + 1);
|
cannam@0
|
141
|
cannam@0
|
142 int distSize = (MAX_RUN_COUNT + 1) * blockSize;
|
cannam@0
|
143
|
cannam@0
|
144 distXSize = blockSize * 2;
|
cannam@0
|
145
|
cannam@0
|
146 // std::cerr << "Matcher::init: distXSize = " << distXSize << std::endl;
|
cannam@0
|
147
|
cannam@0
|
148 distance = (unsigned char **)malloc(distXSize * sizeof(unsigned char *));
|
cannam@0
|
149 bestPathCost = (int **)malloc(distXSize * sizeof(int *));
|
cannam@0
|
150 distYSizes = (int *)malloc(distXSize * sizeof(int));
|
cannam@0
|
151
|
cannam@0
|
152 for (int i = 0; i < blockSize; ++i) {
|
cannam@0
|
153 distance[i] = (unsigned char *)malloc(distSize * sizeof(unsigned char));
|
cannam@0
|
154 bestPathCost[i] = (int *)malloc(distSize * sizeof(int));
|
cannam@0
|
155 distYSizes[i] = distSize;
|
cannam@0
|
156 }
|
cannam@0
|
157 for (int i = blockSize; i < distXSize; ++i) {
|
cannam@0
|
158 distance[i] = 0;
|
cannam@0
|
159 }
|
cannam@0
|
160
|
cannam@0
|
161 first = (int *)malloc(distXSize * sizeof(int));
|
cannam@0
|
162 last = (int *)malloc(distXSize * sizeof(int));
|
cannam@0
|
163
|
cannam@0
|
164 frameCount = 0;
|
cannam@0
|
165 runCount = 0;
|
cannam@0
|
166 // frameRMS = 0;
|
cannam@0
|
167 ltAverage = 0;
|
cannam@0
|
168
|
cannam@0
|
169 if (!silent) print();
|
cannam@0
|
170 } // init
|
cannam@0
|
171
|
cannam@0
|
172 void
|
cannam@0
|
173 Matcher::makeFreqMap(int fftSize, float sampleRate)
|
cannam@0
|
174 {
|
cannam@0
|
175 initVector<int>(freqMap, fftSize/2 + 1);
|
cannam@0
|
176 if (useChromaFrequencyMap)
|
cannam@0
|
177 makeChromaFrequencyMap(fftSize, sampleRate);
|
cannam@0
|
178 else
|
cannam@0
|
179 makeStandardFrequencyMap(fftSize, sampleRate);
|
cannam@0
|
180 } // makeFreqMap()
|
cannam@0
|
181
|
cannam@0
|
182 void
|
cannam@0
|
183 Matcher::makeStandardFrequencyMap(int fftSize, float sampleRate)
|
cannam@0
|
184 {
|
cannam@0
|
185 double binWidth = sampleRate / fftSize;
|
cannam@0
|
186 int crossoverBin = (int)(2 / (pow(2, 1/12.0) - 1));
|
cannam@7
|
187 int crossoverMidi = lrint(log(crossoverBin*binWidth/440.0)/
|
cannam@7
|
188 log(2.0) * 12 + 69);
|
cannam@0
|
189 // freq = 440 * Math.pow(2, (midi-69)/12.0) / binWidth;
|
cannam@0
|
190 int i = 0;
|
cannam@0
|
191 while (i <= crossoverBin) {
|
cannam@0
|
192 freqMap[i] = i;
|
cannam@0
|
193 ++i;
|
cannam@0
|
194 }
|
cannam@0
|
195 while (i <= fftSize/2) {
|
cannam@7
|
196 double midi = log(i*binWidth/440.0) / log(2.0) * 12 + 69;
|
cannam@0
|
197 if (midi > 127)
|
cannam@0
|
198 midi = 127;
|
cannam@0
|
199 freqMap[i++] = crossoverBin + lrint(midi) - crossoverMidi;
|
cannam@0
|
200 }
|
cannam@0
|
201 freqMapSize = freqMap[i-1] + 1;
|
cannam@0
|
202 if (!silent) {
|
cannam@0
|
203 cerr << "Standard map size: " << freqMapSize
|
cannam@0
|
204 << "; Crossover at: " << crossoverBin << endl;
|
cannam@0
|
205 //!!! for (i = 0; i < fftSize / 2; i++)
|
cannam@0
|
206 // cerr << "freqMap[" << i << "] = " << freqMap[i] << endl;
|
cannam@0
|
207 }
|
cannam@0
|
208 } // makeStandardFrequencyMap()
|
cannam@0
|
209
|
cannam@0
|
210 void
|
cannam@0
|
211 Matcher::makeChromaFrequencyMap(int fftSize, float sampleRate)
|
cannam@0
|
212 {
|
cannam@0
|
213 double binWidth = sampleRate / fftSize;
|
cannam@0
|
214 int crossoverBin = (int)(1 / (pow(2, 1/12.0) - 1));
|
cannam@0
|
215 // freq = 440 * Math.pow(2, (midi-69)/12.0) / binWidth;
|
cannam@0
|
216 int i = 0;
|
cannam@0
|
217 while (i <= crossoverBin)
|
cannam@0
|
218 freqMap[i++] = 0;
|
cannam@0
|
219 while (i <= fftSize/2) {
|
cannam@7
|
220 double midi = log(i*binWidth/440.0) / log(2.0) * 12 + 69;
|
cannam@0
|
221 freqMap[i++] = (lrint(midi)) % 12 + 1;
|
cannam@0
|
222 }
|
cannam@0
|
223 freqMapSize = 13;
|
cannam@0
|
224 if (!silent) {
|
cannam@0
|
225 cerr << "Chroma map size: " << freqMapSize
|
cannam@0
|
226 << "; Crossover at: " << crossoverBin << endl;
|
cannam@0
|
227 for (i = 0; i < fftSize / 2; i++)
|
cannam@0
|
228 cerr << "freqMap[" << i << "] = " << freqMap[i] << endl;
|
cannam@0
|
229 }
|
cannam@0
|
230 } // makeChromaFrequencyMap()
|
cannam@0
|
231
|
cannam@0
|
232 void
|
cannam@0
|
233 Matcher::processFrame(double *reBuffer, double *imBuffer)
|
cannam@0
|
234 {
|
cannam@0
|
235 if (!initialised) init();
|
cannam@0
|
236
|
cannam@0
|
237 for (int i = 0; i < (int)newFrame.size(); ++i) {
|
cannam@0
|
238 newFrame[i] = 0;
|
cannam@0
|
239 }
|
cannam@0
|
240 double rms = 0;
|
cannam@0
|
241 for (int i = 0; i <= fftSize/2; i++) {
|
cannam@0
|
242 double mag = reBuffer[i] * reBuffer[i] +
|
cannam@0
|
243 imBuffer[i] * imBuffer[i];
|
cannam@0
|
244 rms += mag;
|
cannam@0
|
245 newFrame[freqMap[i]] += mag;
|
cannam@0
|
246 }
|
cannam@0
|
247 rms = sqrt(rms / (fftSize/2));
|
cannam@0
|
248
|
cannam@0
|
249 int frameIndex = frameCount % blockSize;
|
cannam@0
|
250
|
cannam@0
|
251 if (frameCount >= distXSize) {
|
cannam@0
|
252 // std::cerr << "Resizing " << distXSize << " -> " << distXSize * 2 << std::endl;
|
cannam@0
|
253 distXSize *= 2;
|
cannam@0
|
254 distance = (unsigned char **)realloc(distance, distXSize * sizeof(unsigned char *));
|
cannam@0
|
255 bestPathCost = (int **)realloc(bestPathCost, distXSize * sizeof(int *));
|
cannam@0
|
256 distYSizes = (int *)realloc(distYSizes, distXSize * sizeof(int));
|
cannam@0
|
257 first = (int *)realloc(first, distXSize * sizeof(int));
|
cannam@0
|
258 last = (int *)realloc(last, distXSize * sizeof(int));
|
cannam@0
|
259
|
cannam@0
|
260 for (int i = distXSize/2; i < distXSize; ++i) {
|
cannam@0
|
261 distance[i] = 0;
|
cannam@0
|
262 }
|
cannam@0
|
263 }
|
cannam@0
|
264
|
cannam@0
|
265 if (firstPM && (frameCount >= blockSize)) {
|
cannam@0
|
266
|
cannam@0
|
267 int len = last[frameCount - blockSize] -
|
cannam@0
|
268 first[frameCount - blockSize];
|
cannam@0
|
269
|
cannam@0
|
270 // We need to copy distance[frameCount-blockSize] to
|
cannam@0
|
271 // distance[frameCount], and then truncate
|
cannam@0
|
272 // distance[frameCount-blockSize] to its first len elements.
|
cannam@0
|
273 // Same for bestPathCost.
|
cannam@0
|
274 /*
|
cannam@4
|
275 std::cerr << "Matcher(" << this << "): moving " << distYSizes[frameCount - blockSize] << " from " << frameCount - blockSize << " to "
|
cannam@0
|
276 << frameCount << ", allocating " << len << " for "
|
cannam@0
|
277 << frameCount - blockSize << std::endl;
|
cannam@0
|
278 */
|
cannam@0
|
279 distance[frameCount] = distance[frameCount - blockSize];
|
cannam@0
|
280
|
cannam@0
|
281 distance[frameCount - blockSize] = (unsigned char *)
|
cannam@0
|
282 malloc(len * sizeof(unsigned char));
|
cannam@0
|
283 for (int i = 0; i < len; ++i) {
|
cannam@0
|
284 distance[frameCount - blockSize][i] =
|
cannam@0
|
285 distance[frameCount][i];
|
cannam@0
|
286 }
|
cannam@0
|
287
|
cannam@0
|
288 bestPathCost[frameCount] = bestPathCost[frameCount - blockSize];
|
cannam@0
|
289
|
cannam@0
|
290 bestPathCost[frameCount - blockSize] = (int *)
|
cannam@0
|
291 malloc(len * sizeof(int));
|
cannam@0
|
292 for (int i = 0; i < len; ++i) {
|
cannam@0
|
293 bestPathCost[frameCount - blockSize][i] =
|
cannam@0
|
294 bestPathCost[frameCount][i];
|
cannam@0
|
295 }
|
cannam@0
|
296
|
cannam@0
|
297 distYSizes[frameCount] = distYSizes[frameCount - blockSize];
|
cannam@0
|
298 distYSizes[frameCount - blockSize] = len;
|
cannam@0
|
299 }
|
cannam@0
|
300
|
cannam@0
|
301 double totalEnergy = 0;
|
cannam@0
|
302 if (useSpectralDifference) {
|
cannam@0
|
303 for (int i = 0; i < freqMapSize; i++) {
|
cannam@0
|
304 totalEnergy += newFrame[i];
|
cannam@0
|
305 if (newFrame[i] > prevFrame[i]) {
|
cannam@0
|
306 frames[frameIndex][i] = newFrame[i] - prevFrame[i];
|
cannam@0
|
307 } else {
|
cannam@0
|
308 frames[frameIndex][i] = 0;
|
cannam@0
|
309 }
|
cannam@0
|
310 }
|
cannam@0
|
311 } else {
|
cannam@0
|
312 for (int i = 0; i < freqMapSize; i++) {
|
cannam@0
|
313 frames[frameIndex][i] = newFrame[i];
|
cannam@0
|
314 totalEnergy += frames[frameIndex][i];
|
cannam@0
|
315 }
|
cannam@0
|
316 }
|
cannam@0
|
317 frames[frameIndex][freqMapSize] = totalEnergy;
|
cannam@0
|
318
|
cannam@0
|
319 double decay = frameCount >= 200 ? 0.99:
|
cannam@0
|
320 (frameCount < 100? 0: (frameCount - 100) / 100.0);
|
cannam@0
|
321
|
cannam@0
|
322 if (ltAverage == 0)
|
cannam@0
|
323 ltAverage = totalEnergy;
|
cannam@0
|
324 else
|
cannam@0
|
325 ltAverage = ltAverage * decay + totalEnergy * (1.0 - decay);
|
cannam@0
|
326
|
cannam@0
|
327 // System.err.println(Format.d(ltAverage,4) + " " +
|
cannam@0
|
328 // Format.d(totalEnergy) + " " +
|
cannam@0
|
329 // Format.d(frameRMS));
|
cannam@0
|
330
|
cannam@0
|
331 // std::cerr << "ltAverage: " << ltAverage << ", totalEnergy: " << totalEnergy << ", frameRMS: " << rms << std::endl;
|
cannam@0
|
332
|
cannam@0
|
333 if (rms <= 0.01) //!!! silenceThreshold)
|
cannam@0
|
334 for (int i = 0; i < freqMapSize; i++)
|
cannam@0
|
335 frames[frameIndex][i] = 0;
|
cannam@0
|
336 else if (normalise1)
|
cannam@0
|
337 for (int i = 0; i < freqMapSize; i++)
|
cannam@0
|
338 frames[frameIndex][i] /= totalEnergy;
|
cannam@0
|
339 else if (normalise3)
|
cannam@0
|
340 for (int i = 0; i < freqMapSize; i++)
|
cannam@0
|
341 frames[frameIndex][i] /= ltAverage;
|
cannam@0
|
342
|
cannam@0
|
343 int stop = otherMatcher->frameCount;
|
cannam@0
|
344 int index = stop - blockSize;
|
cannam@0
|
345 if (index < 0)
|
cannam@0
|
346 index = 0;
|
cannam@0
|
347 first[frameCount] = index;
|
cannam@0
|
348 last[frameCount] = stop;
|
cannam@0
|
349
|
cannam@0
|
350 bool overflow = false;
|
cannam@0
|
351 int mn= -1;
|
cannam@0
|
352 int mx= -1;
|
cannam@0
|
353 for ( ; index < stop; index++) {
|
cannam@0
|
354 int dMN = calcDistance(frames[frameIndex],
|
cannam@0
|
355 otherMatcher->frames[index % blockSize]);
|
cannam@0
|
356 if (mx<0)
|
cannam@0
|
357 mx = mn = dMN;
|
cannam@0
|
358 else if (dMN > mx)
|
cannam@0
|
359 mx = dMN;
|
cannam@0
|
360 else if (dMN < mn)
|
cannam@0
|
361 mn = dMN;
|
cannam@0
|
362 if (dMN >= 255) {
|
cannam@0
|
363 overflow = true;
|
cannam@0
|
364 dMN = 255;
|
cannam@0
|
365 }
|
cannam@0
|
366 if ((frameCount == 0) && (index == 0)) // first element
|
cannam@0
|
367 setValue(0, 0, 0, 0, dMN);
|
cannam@0
|
368 else if (frameCount == 0) // first row
|
cannam@0
|
369 setValue(0, index, ADVANCE_OTHER,
|
cannam@0
|
370 getValue(0, index-1, true), dMN);
|
cannam@0
|
371 else if (index == 0) // first column
|
cannam@0
|
372 setValue(frameCount, index, ADVANCE_THIS,
|
cannam@0
|
373 getValue(frameCount - 1, 0, true), dMN);
|
cannam@0
|
374 else if (index == otherMatcher->frameCount - blockSize) {
|
cannam@0
|
375 // missing value(s) due to cutoff
|
cannam@0
|
376 // - no previous value in current row (resp. column)
|
cannam@0
|
377 // - no diagonal value if prev. dir. == curr. dirn
|
cannam@0
|
378 int min2 = getValue(frameCount - 1, index, true);
|
cannam@0
|
379 // if ((firstPM && (first[frameCount - 1] == index)) ||
|
cannam@0
|
380 // (!firstPM && (last[index-1] < frameCount)))
|
cannam@0
|
381 if (first[frameCount - 1] == index)
|
cannam@0
|
382 setValue(frameCount, index, ADVANCE_THIS, min2, dMN);
|
cannam@0
|
383 else {
|
cannam@0
|
384 int min1 = getValue(frameCount - 1, index - 1, true);
|
cannam@0
|
385 if (min1 + dMN <= min2)
|
cannam@0
|
386 setValue(frameCount, index, ADVANCE_BOTH, min1,dMN);
|
cannam@0
|
387 else
|
cannam@0
|
388 setValue(frameCount, index, ADVANCE_THIS, min2,dMN);
|
cannam@0
|
389 }
|
cannam@0
|
390 } else {
|
cannam@0
|
391 int min1 = getValue(frameCount, index-1, true);
|
cannam@0
|
392 int min2 = getValue(frameCount - 1, index, true);
|
cannam@0
|
393 int min3 = getValue(frameCount - 1, index-1, true);
|
cannam@0
|
394 if (min1 <= min2) {
|
cannam@0
|
395 if (min3 + dMN <= min1)
|
cannam@0
|
396 setValue(frameCount, index, ADVANCE_BOTH, min3,dMN);
|
cannam@0
|
397 else
|
cannam@0
|
398 setValue(frameCount, index, ADVANCE_OTHER,min1,dMN);
|
cannam@0
|
399 } else {
|
cannam@0
|
400 if (min3 + dMN <= min2)
|
cannam@0
|
401 setValue(frameCount, index, ADVANCE_BOTH, min3,dMN);
|
cannam@0
|
402 else
|
cannam@0
|
403 setValue(frameCount, index, ADVANCE_THIS, min2,dMN);
|
cannam@0
|
404 }
|
cannam@0
|
405 }
|
cannam@0
|
406 otherMatcher->last[index]++;
|
cannam@0
|
407 } // loop for row (resp. column)
|
cannam@0
|
408
|
cannam@0
|
409 vector<double> tmp = prevFrame;
|
cannam@0
|
410 prevFrame = newFrame;
|
cannam@0
|
411 newFrame = tmp;
|
cannam@0
|
412
|
cannam@0
|
413 frameCount++;
|
cannam@0
|
414 runCount++;
|
cannam@0
|
415
|
cannam@0
|
416 otherMatcher->runCount = 0;
|
cannam@0
|
417
|
cannam@0
|
418 if (overflow && !silent)
|
cannam@0
|
419 cerr << "WARNING: overflow in distance metric: "
|
cannam@0
|
420 << "frame " << frameCount << ", val = " << mx << endl;
|
cannam@0
|
421
|
cannam@0
|
422 if (!silent)
|
cannam@0
|
423 std::cerr << "Frame " << frameCount << ", d = " << (mx-mn) << std::endl;
|
cannam@0
|
424
|
cannam@0
|
425 if ((frameCount % 100) == 0) {
|
cannam@0
|
426 if (!silent) {
|
cannam@0
|
427 cerr << "Progress:" << frameCount << " " << ltAverage << endl;
|
cannam@0
|
428 // Profile.report();
|
cannam@0
|
429 }
|
cannam@0
|
430 }
|
cannam@0
|
431 //!!! if (frameCount == maxFrames)
|
cannam@0
|
432 // closeStreams();
|
cannam@0
|
433 // }
|
cannam@0
|
434 } // processFrame()
|
cannam@0
|
435
|
cannam@0
|
436 int
|
cannam@0
|
437 Matcher::calcDistance(const vector<double> &f1, const vector<double> &f2)
|
cannam@0
|
438 {
|
cannam@0
|
439 double d = 0;
|
cannam@0
|
440 double sum = 0;
|
cannam@0
|
441 for (int i = 0; i < freqMapSize; i++) {
|
cannam@0
|
442 d += fabs(f1[i] - f2[i]);
|
cannam@0
|
443 sum += f1[i] + f2[i];
|
cannam@0
|
444 }
|
cannam@0
|
445 // System.err.print(" " + Format.d(d,3));
|
cannam@0
|
446 if (sum == 0)
|
cannam@0
|
447 return 0;
|
cannam@0
|
448 if (normalise2)
|
cannam@0
|
449 return (int)(scale * d / sum); // 0 <= d/sum <= 2
|
cannam@0
|
450 if (!normalise4)
|
cannam@0
|
451 return (int)(scale * d);
|
cannam@0
|
452 // double weight = (5 + Math.log(f1[freqMapSize] + f2[freqMapSize]))/10.0;
|
cannam@0
|
453 double weight = (8 + log(sum)) / 10.0;
|
cannam@0
|
454 // if (weight < mins) {
|
cannam@0
|
455 // mins = weight;
|
cannam@0
|
456 // System.err.println(Format.d(mins,3) + " " + Format.d(maxs));
|
cannam@0
|
457 // }
|
cannam@0
|
458 // if (weight > maxs) {
|
cannam@0
|
459 // maxs = weight;
|
cannam@0
|
460 // System.err.println(Format.d(mins,3) + " " + Format.d(maxs));
|
cannam@0
|
461 // }
|
cannam@0
|
462 if (weight < 0)
|
cannam@0
|
463 weight = 0;
|
cannam@0
|
464 else if (weight > 1)
|
cannam@0
|
465 weight = 1;
|
cannam@0
|
466 return (int)(scale * d / sum * weight);
|
cannam@0
|
467 } // calcDistance()
|
cannam@0
|
468
|
cannam@0
|
469 int
|
cannam@0
|
470 Matcher::getValue(int i, int j, bool firstAttempt)
|
cannam@0
|
471 {
|
cannam@0
|
472 if (firstPM)
|
cannam@0
|
473 return bestPathCost[i][j - first[i]];
|
cannam@0
|
474 else
|
cannam@0
|
475 return otherMatcher->bestPathCost[j][i - otherMatcher->first[j]];
|
cannam@0
|
476 } // getValue()
|
cannam@0
|
477
|
cannam@0
|
478 void
|
cannam@0
|
479 Matcher::setValue(int i, int j, int dir, int value, int dMN)
|
cannam@0
|
480 {
|
cannam@0
|
481 if (firstPM) {
|
cannam@0
|
482 distance[i][j - first[i]] = (unsigned char)((dMN & MASK) | dir);
|
cannam@0
|
483 bestPathCost[i][j - first[i]] =
|
cannam@0
|
484 (value + (dir==ADVANCE_BOTH? dMN*2: dMN));
|
cannam@0
|
485 } else {
|
cannam@0
|
486 if (dir == ADVANCE_THIS)
|
cannam@0
|
487 dir = ADVANCE_OTHER;
|
cannam@0
|
488 else if (dir == ADVANCE_OTHER)
|
cannam@0
|
489 dir = ADVANCE_THIS;
|
cannam@0
|
490 int idx = i - otherMatcher->first[j];
|
cannam@0
|
491 if (idx == (int)otherMatcher->distYSizes[j]) {
|
cannam@0
|
492 // This should never happen, but if we allow arbitrary
|
cannam@0
|
493 // pauses in either direction, and arbitrary lengths at
|
cannam@0
|
494 // end, it is better than a segmentation fault.
|
cannam@0
|
495 std::cerr << "Emergency resize: " << idx << " -> " << idx * 2 << std::endl;
|
cannam@0
|
496 otherMatcher->distYSizes[j] = idx * 2;
|
cannam@0
|
497 otherMatcher->bestPathCost[j] =
|
cannam@0
|
498 (int *)realloc(otherMatcher->bestPathCost[j],
|
cannam@0
|
499 idx * 2 * sizeof(int));
|
cannam@0
|
500 otherMatcher->distance[j] =
|
cannam@0
|
501 (unsigned char *)realloc(otherMatcher->distance[j],
|
cannam@0
|
502 idx * 2 * sizeof(unsigned char));
|
cannam@0
|
503 }
|
cannam@0
|
504 otherMatcher->distance[j][idx] = (unsigned char)((dMN & MASK) | dir);
|
cannam@0
|
505 otherMatcher->bestPathCost[j][idx] =
|
cannam@0
|
506 (value + (dir==ADVANCE_BOTH? dMN*2: dMN));
|
cannam@0
|
507 }
|
cannam@0
|
508 } // setValue()
|
cannam@0
|
509
|