Chris@3
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
Chris@3
|
2 /*
|
Chris@3
|
3 Permission is hereby granted, free of charge, to any person
|
Chris@3
|
4 obtaining a copy of this software and associated documentation
|
Chris@3
|
5 files (the "Software"), to deal in the Software without
|
Chris@3
|
6 restriction, including without limitation the rights to use, copy,
|
Chris@3
|
7 modify, merge, publish, distribute, sublicense, and/or sell copies
|
Chris@3
|
8 of the Software, and to permit persons to whom the Software is
|
Chris@3
|
9 furnished to do so, subject to the following conditions:
|
Chris@3
|
10
|
Chris@3
|
11 The above copyright notice and this permission notice shall be
|
Chris@3
|
12 included in all copies or substantial portions of the Software.
|
Chris@3
|
13
|
Chris@3
|
14 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
Chris@3
|
15 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
Chris@3
|
16 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
Chris@3
|
17 NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
|
Chris@3
|
18 ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
|
Chris@3
|
19 CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
|
Chris@3
|
20 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
|
Chris@3
|
21 */
|
Chris@3
|
22
|
Chris@3
|
23 #include "CepstrumPitchTracker.h"
|
Chris@3
|
24
|
Chris@3
|
25 #include <vector>
|
Chris@3
|
26 #include <algorithm>
|
Chris@3
|
27
|
Chris@3
|
28 #include <cstdio>
|
Chris@3
|
29 #include <cmath>
|
Chris@3
|
30 #include <complex>
|
Chris@3
|
31
|
Chris@3
|
32 using std::string;
|
Chris@3
|
33
|
Chris@3
|
34 CepstrumPitchTracker::CepstrumPitchTracker(float inputSampleRate) :
|
Chris@3
|
35 Plugin(inputSampleRate),
|
Chris@3
|
36 m_channels(0),
|
Chris@3
|
37 m_stepSize(256),
|
Chris@3
|
38 m_blockSize(1024),
|
Chris@3
|
39 m_fmin(50),
|
Chris@3
|
40 m_fmax(1000),
|
Chris@3
|
41 m_histlen(3),
|
Chris@3
|
42 m_binFrom(0),
|
Chris@3
|
43 m_binTo(0),
|
Chris@3
|
44 m_bins(0),
|
Chris@3
|
45 m_history(0)
|
Chris@3
|
46 {
|
Chris@3
|
47 }
|
Chris@3
|
48
|
Chris@3
|
49 CepstrumPitchTracker::~CepstrumPitchTracker()
|
Chris@3
|
50 {
|
Chris@3
|
51 if (m_history) {
|
Chris@3
|
52 for (int i = 0; i < m_histlen; ++i) {
|
Chris@3
|
53 delete[] m_history[i];
|
Chris@3
|
54 }
|
Chris@3
|
55 delete[] m_history;
|
Chris@3
|
56 }
|
Chris@3
|
57 }
|
Chris@3
|
58
|
Chris@3
|
59 string
|
Chris@3
|
60 CepstrumPitchTracker::getIdentifier() const
|
Chris@3
|
61 {
|
Chris@3
|
62 return "cepstrum-pitch";
|
Chris@3
|
63 }
|
Chris@3
|
64
|
Chris@3
|
65 string
|
Chris@3
|
66 CepstrumPitchTracker::getName() const
|
Chris@3
|
67 {
|
Chris@3
|
68 return "Cepstrum Pitch Tracker";
|
Chris@3
|
69 }
|
Chris@3
|
70
|
Chris@3
|
71 string
|
Chris@3
|
72 CepstrumPitchTracker::getDescription() const
|
Chris@3
|
73 {
|
Chris@3
|
74 return "Estimate f0 of monophonic material using a cepstrum method.";
|
Chris@3
|
75 }
|
Chris@3
|
76
|
Chris@3
|
77 string
|
Chris@3
|
78 CepstrumPitchTracker::getMaker() const
|
Chris@3
|
79 {
|
Chris@3
|
80 return "Chris Cannam";
|
Chris@3
|
81 }
|
Chris@3
|
82
|
Chris@3
|
83 int
|
Chris@3
|
84 CepstrumPitchTracker::getPluginVersion() const
|
Chris@3
|
85 {
|
Chris@3
|
86 // Increment this each time you release a version that behaves
|
Chris@3
|
87 // differently from the previous one
|
Chris@3
|
88 return 1;
|
Chris@3
|
89 }
|
Chris@3
|
90
|
Chris@3
|
91 string
|
Chris@3
|
92 CepstrumPitchTracker::getCopyright() const
|
Chris@3
|
93 {
|
Chris@3
|
94 return "Freely redistributable (BSD license)";
|
Chris@3
|
95 }
|
Chris@3
|
96
|
Chris@3
|
97 CepstrumPitchTracker::InputDomain
|
Chris@3
|
98 CepstrumPitchTracker::getInputDomain() const
|
Chris@3
|
99 {
|
Chris@3
|
100 return FrequencyDomain;
|
Chris@3
|
101 }
|
Chris@3
|
102
|
Chris@3
|
103 size_t
|
Chris@3
|
104 CepstrumPitchTracker::getPreferredBlockSize() const
|
Chris@3
|
105 {
|
Chris@3
|
106 return 1024;
|
Chris@3
|
107 }
|
Chris@3
|
108
|
Chris@3
|
109 size_t
|
Chris@3
|
110 CepstrumPitchTracker::getPreferredStepSize() const
|
Chris@3
|
111 {
|
Chris@3
|
112 return 256;
|
Chris@3
|
113 }
|
Chris@3
|
114
|
Chris@3
|
115 size_t
|
Chris@3
|
116 CepstrumPitchTracker::getMinChannelCount() const
|
Chris@3
|
117 {
|
Chris@3
|
118 return 1;
|
Chris@3
|
119 }
|
Chris@3
|
120
|
Chris@3
|
121 size_t
|
Chris@3
|
122 CepstrumPitchTracker::getMaxChannelCount() const
|
Chris@3
|
123 {
|
Chris@3
|
124 return 1;
|
Chris@3
|
125 }
|
Chris@3
|
126
|
Chris@3
|
127 CepstrumPitchTracker::ParameterList
|
Chris@3
|
128 CepstrumPitchTracker::getParameterDescriptors() const
|
Chris@3
|
129 {
|
Chris@3
|
130 ParameterList list;
|
Chris@3
|
131 return list;
|
Chris@3
|
132 }
|
Chris@3
|
133
|
Chris@3
|
134 float
|
Chris@3
|
135 CepstrumPitchTracker::getParameter(string identifier) const
|
Chris@3
|
136 {
|
Chris@3
|
137 return 0.f;
|
Chris@3
|
138 }
|
Chris@3
|
139
|
Chris@3
|
140 void
|
Chris@3
|
141 CepstrumPitchTracker::setParameter(string identifier, float value)
|
Chris@3
|
142 {
|
Chris@3
|
143 }
|
Chris@3
|
144
|
Chris@3
|
145 CepstrumPitchTracker::ProgramList
|
Chris@3
|
146 CepstrumPitchTracker::getPrograms() const
|
Chris@3
|
147 {
|
Chris@3
|
148 ProgramList list;
|
Chris@3
|
149 return list;
|
Chris@3
|
150 }
|
Chris@3
|
151
|
Chris@3
|
152 string
|
Chris@3
|
153 CepstrumPitchTracker::getCurrentProgram() const
|
Chris@3
|
154 {
|
Chris@3
|
155 return ""; // no programs
|
Chris@3
|
156 }
|
Chris@3
|
157
|
Chris@3
|
158 void
|
Chris@3
|
159 CepstrumPitchTracker::selectProgram(string name)
|
Chris@3
|
160 {
|
Chris@3
|
161 }
|
Chris@3
|
162
|
Chris@3
|
163 CepstrumPitchTracker::OutputList
|
Chris@3
|
164 CepstrumPitchTracker::getOutputDescriptors() const
|
Chris@3
|
165 {
|
Chris@3
|
166 OutputList outputs;
|
Chris@3
|
167
|
Chris@3
|
168 int n = 0;
|
Chris@3
|
169
|
Chris@3
|
170 OutputDescriptor d;
|
Chris@3
|
171
|
Chris@3
|
172 d.identifier = "f0";
|
Chris@3
|
173 d.name = "Estimated f0";
|
Chris@3
|
174 d.description = "Estimated fundamental frequency";
|
Chris@3
|
175 d.unit = "Hz";
|
Chris@3
|
176 d.hasFixedBinCount = true;
|
Chris@3
|
177 d.binCount = 1;
|
Chris@3
|
178 d.hasKnownExtents = true;
|
Chris@3
|
179 d.minValue = m_fmin;
|
Chris@3
|
180 d.maxValue = m_fmax;
|
Chris@3
|
181 d.isQuantized = false;
|
Chris@3
|
182 d.sampleType = OutputDescriptor::FixedSampleRate;
|
Chris@3
|
183 d.sampleRate = (m_inputSampleRate / m_stepSize);
|
Chris@3
|
184 d.hasDuration = false;
|
Chris@3
|
185 outputs.push_back(d);
|
Chris@3
|
186
|
Chris@3
|
187 return outputs;
|
Chris@3
|
188 }
|
Chris@3
|
189
|
Chris@3
|
190 bool
|
Chris@3
|
191 CepstrumPitchTracker::initialise(size_t channels, size_t stepSize, size_t blockSize)
|
Chris@3
|
192 {
|
Chris@3
|
193 if (channels < getMinChannelCount() ||
|
Chris@3
|
194 channels > getMaxChannelCount()) return false;
|
Chris@3
|
195
|
Chris@3
|
196 // std::cerr << "CepstrumPitchTracker::initialise: channels = " << channels
|
Chris@3
|
197 // << ", stepSize = " << stepSize << ", blockSize = " << blockSize
|
Chris@3
|
198 // << std::endl;
|
Chris@3
|
199
|
Chris@3
|
200 m_channels = channels;
|
Chris@3
|
201 m_stepSize = stepSize;
|
Chris@3
|
202 m_blockSize = blockSize;
|
Chris@3
|
203
|
Chris@3
|
204 m_binFrom = int(m_inputSampleRate / m_fmax);
|
Chris@3
|
205 m_binTo = int(m_inputSampleRate / m_fmin);
|
Chris@3
|
206
|
Chris@3
|
207 if (m_binTo >= (int)m_blockSize / 2) {
|
Chris@3
|
208 m_binTo = m_blockSize / 2 - 1;
|
Chris@3
|
209 }
|
Chris@3
|
210
|
Chris@3
|
211 m_bins = (m_binTo - m_binFrom) + 1;
|
Chris@3
|
212
|
Chris@3
|
213 m_history = new double *[m_histlen];
|
Chris@3
|
214 for (int i = 0; i < m_histlen; ++i) {
|
Chris@3
|
215 m_history[i] = new double[m_bins];
|
Chris@3
|
216 }
|
Chris@3
|
217
|
Chris@3
|
218 reset();
|
Chris@3
|
219
|
Chris@3
|
220 return true;
|
Chris@3
|
221 }
|
Chris@3
|
222
|
Chris@3
|
223 void
|
Chris@3
|
224 CepstrumPitchTracker::reset()
|
Chris@3
|
225 {
|
Chris@3
|
226 for (int i = 0; i < m_histlen; ++i) {
|
Chris@3
|
227 for (int j = 0; j < m_bins; ++j) {
|
Chris@3
|
228 m_history[i][j] = 0.0;
|
Chris@3
|
229 }
|
Chris@3
|
230 }
|
Chris@3
|
231 }
|
Chris@3
|
232
|
Chris@3
|
233 void
|
Chris@3
|
234 CepstrumPitchTracker::filter(const double *cep, double *result)
|
Chris@3
|
235 {
|
Chris@3
|
236 int hix = m_histlen - 1; // current history index
|
Chris@3
|
237
|
Chris@3
|
238 // roll back the history
|
Chris@3
|
239 if (m_histlen > 1) {
|
Chris@3
|
240 double *oldest = m_history[0];
|
Chris@3
|
241 for (int i = 1; i < m_histlen; ++i) {
|
Chris@3
|
242 m_history[i-1] = m_history[i];
|
Chris@3
|
243 }
|
Chris@3
|
244 // and stick this back in the newest spot, to recycle
|
Chris@3
|
245 m_history[hix] = oldest;
|
Chris@3
|
246 }
|
Chris@3
|
247
|
Chris@3
|
248 for (int i = 0; i < m_bins; ++i) {
|
Chris@3
|
249 m_history[hix][i] = cep[i + m_binFrom];
|
Chris@3
|
250 }
|
Chris@3
|
251
|
Chris@3
|
252 for (int i = 0; i < m_bins; ++i) {
|
Chris@3
|
253 double mean = 0.0;
|
Chris@3
|
254 for (int j = 0; j < m_histlen; ++j) {
|
Chris@3
|
255 mean += m_history[j][i];
|
Chris@3
|
256 }
|
Chris@3
|
257 mean /= m_histlen;
|
Chris@3
|
258 result[i] = mean;
|
Chris@3
|
259 }
|
Chris@3
|
260 }
|
Chris@3
|
261
|
Chris@3
|
262 CepstrumPitchTracker::FeatureSet
|
Chris@3
|
263 CepstrumPitchTracker::process(const float *const *inputBuffers, Vamp::RealTime timestamp)
|
Chris@3
|
264 {
|
Chris@3
|
265 FeatureSet fs;
|
Chris@3
|
266
|
Chris@3
|
267 int bs = m_blockSize;
|
Chris@3
|
268 int hs = m_blockSize/2 + 1;
|
Chris@3
|
269
|
Chris@3
|
270 double *rawcep = new double[bs];
|
Chris@3
|
271 double *io = new double[bs];
|
Chris@3
|
272 double *logmag = new double[bs];
|
Chris@3
|
273
|
Chris@3
|
274 // The "forward difference" method
|
Chris@3
|
275
|
Chris@3
|
276 for (int i = 0; i < hs; ++i) {
|
Chris@3
|
277
|
Chris@3
|
278 double power =
|
Chris@3
|
279 inputBuffers[0][i*2 ] * inputBuffers[0][i*2 ] +
|
Chris@3
|
280 inputBuffers[0][i*2+1] * inputBuffers[0][i*2+1];
|
Chris@3
|
281 double mag = sqrt(power);
|
Chris@3
|
282
|
Chris@3
|
283 double lm = log(mag + 0.00000001);
|
Chris@3
|
284
|
Chris@3
|
285 logmag[bs/2 + i - 1] = lm;
|
Chris@3
|
286 if (i < hs-1) {
|
Chris@3
|
287 logmag[bs/2 - i - 1] = lm;
|
Chris@3
|
288 }
|
Chris@3
|
289 }
|
Chris@3
|
290
|
Chris@3
|
291 fft(bs, false, logmag, 0, rawcep, io);
|
Chris@3
|
292
|
Chris@3
|
293 for (int i = 0; i < hs; ++i) {
|
Chris@3
|
294 rawcep[i] = fabs(io[i]) - fabs(rawcep[i]);
|
Chris@3
|
295 }
|
Chris@3
|
296
|
Chris@3
|
297 delete[] logmag;
|
Chris@3
|
298 delete[] io;
|
Chris@3
|
299
|
Chris@3
|
300 int n = m_bins;
|
Chris@3
|
301 double *data = new double[n];
|
Chris@3
|
302 filter(rawcep, data);
|
Chris@3
|
303 delete[] rawcep;
|
Chris@3
|
304
|
Chris@3
|
305 double maxval = 0.0;
|
Chris@3
|
306 int maxbin = 0;
|
Chris@3
|
307 double abstot = 0.0;
|
Chris@3
|
308
|
Chris@3
|
309 for (int i = 0; i < n; ++i) {
|
Chris@3
|
310 if (data[i] > maxval) {
|
Chris@3
|
311 maxval = data[i];
|
Chris@3
|
312 maxbin = i;
|
Chris@3
|
313 }
|
Chris@3
|
314 abstot += fabs(data[i]);
|
Chris@3
|
315 }
|
Chris@3
|
316
|
Chris@3
|
317 double aroundPeak = 0.0;
|
Chris@3
|
318 double peakProportion = 0.0;
|
Chris@3
|
319 if (maxval > 0.0) {
|
Chris@3
|
320 aroundPeak += fabs(maxval);
|
Chris@3
|
321 int i = maxbin - 1;
|
Chris@3
|
322 while (i > 0 && data[i] <= data[i+1]) {
|
Chris@3
|
323 aroundPeak += fabs(data[i]);
|
Chris@3
|
324 --i;
|
Chris@3
|
325 }
|
Chris@3
|
326 i = maxbin + 1;
|
Chris@3
|
327 while (i < n && data[i] <= data[i-1]) {
|
Chris@3
|
328 aroundPeak += fabs(data[i]);
|
Chris@3
|
329 ++i;
|
Chris@3
|
330 }
|
Chris@3
|
331 }
|
Chris@3
|
332 peakProportion = aroundPeak / abstot;
|
Chris@3
|
333
|
Chris@3
|
334 // std::cerr << "peakProportion = " << peakProportion << std::endl;
|
Chris@3
|
335 // std::cerr << "peak = " << m_inputSampleRate / (maxbin + m_binFrom) << std::endl;
|
Chris@3
|
336
|
Chris@3
|
337 if (peakProportion >= 0.03) {
|
Chris@3
|
338 Feature f;
|
Chris@3
|
339 f.hasTimestamp = true;
|
Chris@3
|
340 f.timestamp = timestamp;
|
Chris@3
|
341 f.values.push_back(m_inputSampleRate / (maxbin + m_binFrom));
|
Chris@3
|
342 fs[0].push_back(f);
|
Chris@3
|
343 }
|
Chris@3
|
344
|
Chris@3
|
345 delete[] data;
|
Chris@3
|
346 return fs;
|
Chris@3
|
347 }
|
Chris@3
|
348
|
Chris@3
|
349 CepstrumPitchTracker::FeatureSet
|
Chris@3
|
350 CepstrumPitchTracker::getRemainingFeatures()
|
Chris@3
|
351 {
|
Chris@3
|
352 FeatureSet fs;
|
Chris@3
|
353 return fs;
|
Chris@3
|
354 }
|
Chris@3
|
355
|
Chris@3
|
356 void
|
Chris@3
|
357 CepstrumPitchTracker::fft(unsigned int n, bool inverse,
|
Chris@3
|
358 double *ri, double *ii, double *ro, double *io)
|
Chris@3
|
359 {
|
Chris@3
|
360 if (!ri || !ro || !io) return;
|
Chris@3
|
361
|
Chris@3
|
362 unsigned int bits;
|
Chris@3
|
363 unsigned int i, j, k, m;
|
Chris@3
|
364 unsigned int blockSize, blockEnd;
|
Chris@3
|
365
|
Chris@3
|
366 double tr, ti;
|
Chris@3
|
367
|
Chris@3
|
368 if (n < 2) return;
|
Chris@3
|
369 if (n & (n-1)) return;
|
Chris@3
|
370
|
Chris@3
|
371 double angle = 2.0 * M_PI;
|
Chris@3
|
372 if (inverse) angle = -angle;
|
Chris@3
|
373
|
Chris@3
|
374 for (i = 0; ; ++i) {
|
Chris@3
|
375 if (n & (1 << i)) {
|
Chris@3
|
376 bits = i;
|
Chris@3
|
377 break;
|
Chris@3
|
378 }
|
Chris@3
|
379 }
|
Chris@3
|
380
|
Chris@3
|
381 static unsigned int tableSize = 0;
|
Chris@3
|
382 static int *table = 0;
|
Chris@3
|
383
|
Chris@3
|
384 if (tableSize != n) {
|
Chris@3
|
385
|
Chris@3
|
386 delete[] table;
|
Chris@3
|
387
|
Chris@3
|
388 table = new int[n];
|
Chris@3
|
389
|
Chris@3
|
390 for (i = 0; i < n; ++i) {
|
Chris@3
|
391
|
Chris@3
|
392 m = i;
|
Chris@3
|
393
|
Chris@3
|
394 for (j = k = 0; j < bits; ++j) {
|
Chris@3
|
395 k = (k << 1) | (m & 1);
|
Chris@3
|
396 m >>= 1;
|
Chris@3
|
397 }
|
Chris@3
|
398
|
Chris@3
|
399 table[i] = k;
|
Chris@3
|
400 }
|
Chris@3
|
401
|
Chris@3
|
402 tableSize = n;
|
Chris@3
|
403 }
|
Chris@3
|
404
|
Chris@3
|
405 if (ii) {
|
Chris@3
|
406 for (i = 0; i < n; ++i) {
|
Chris@3
|
407 ro[table[i]] = ri[i];
|
Chris@3
|
408 io[table[i]] = ii[i];
|
Chris@3
|
409 }
|
Chris@3
|
410 } else {
|
Chris@3
|
411 for (i = 0; i < n; ++i) {
|
Chris@3
|
412 ro[table[i]] = ri[i];
|
Chris@3
|
413 io[table[i]] = 0.0;
|
Chris@3
|
414 }
|
Chris@3
|
415 }
|
Chris@3
|
416
|
Chris@3
|
417 blockEnd = 1;
|
Chris@3
|
418
|
Chris@3
|
419 for (blockSize = 2; blockSize <= n; blockSize <<= 1) {
|
Chris@3
|
420
|
Chris@3
|
421 double delta = angle / (double)blockSize;
|
Chris@3
|
422 double sm2 = -sin(-2 * delta);
|
Chris@3
|
423 double sm1 = -sin(-delta);
|
Chris@3
|
424 double cm2 = cos(-2 * delta);
|
Chris@3
|
425 double cm1 = cos(-delta);
|
Chris@3
|
426 double w = 2 * cm1;
|
Chris@3
|
427 double ar[3], ai[3];
|
Chris@3
|
428
|
Chris@3
|
429 for (i = 0; i < n; i += blockSize) {
|
Chris@3
|
430
|
Chris@3
|
431 ar[2] = cm2;
|
Chris@3
|
432 ar[1] = cm1;
|
Chris@3
|
433
|
Chris@3
|
434 ai[2] = sm2;
|
Chris@3
|
435 ai[1] = sm1;
|
Chris@3
|
436
|
Chris@3
|
437 for (j = i, m = 0; m < blockEnd; j++, m++) {
|
Chris@3
|
438
|
Chris@3
|
439 ar[0] = w * ar[1] - ar[2];
|
Chris@3
|
440 ar[2] = ar[1];
|
Chris@3
|
441 ar[1] = ar[0];
|
Chris@3
|
442
|
Chris@3
|
443 ai[0] = w * ai[1] - ai[2];
|
Chris@3
|
444 ai[2] = ai[1];
|
Chris@3
|
445 ai[1] = ai[0];
|
Chris@3
|
446
|
Chris@3
|
447 k = j + blockEnd;
|
Chris@3
|
448 tr = ar[0] * ro[k] - ai[0] * io[k];
|
Chris@3
|
449 ti = ar[0] * io[k] + ai[0] * ro[k];
|
Chris@3
|
450
|
Chris@3
|
451 ro[k] = ro[j] - tr;
|
Chris@3
|
452 io[k] = io[j] - ti;
|
Chris@3
|
453
|
Chris@3
|
454 ro[j] += tr;
|
Chris@3
|
455 io[j] += ti;
|
Chris@3
|
456 }
|
Chris@3
|
457 }
|
Chris@3
|
458
|
Chris@3
|
459 blockEnd = blockSize;
|
Chris@3
|
460 }
|
Chris@3
|
461 }
|
Chris@3
|
462
|
Chris@3
|
463
|