Chris@3
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
Chris@3
|
2 /*
|
Chris@31
|
3 This file is Copyright (c) 2012 Chris Cannam
|
Chris@31
|
4
|
Chris@3
|
5 Permission is hereby granted, free of charge, to any person
|
Chris@3
|
6 obtaining a copy of this software and associated documentation
|
Chris@3
|
7 files (the "Software"), to deal in the Software without
|
Chris@3
|
8 restriction, including without limitation the rights to use, copy,
|
Chris@3
|
9 modify, merge, publish, distribute, sublicense, and/or sell copies
|
Chris@3
|
10 of the Software, and to permit persons to whom the Software is
|
Chris@3
|
11 furnished to do so, subject to the following conditions:
|
Chris@3
|
12
|
Chris@3
|
13 The above copyright notice and this permission notice shall be
|
Chris@3
|
14 included in all copies or substantial portions of the Software.
|
Chris@3
|
15
|
Chris@3
|
16 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
Chris@3
|
17 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
Chris@3
|
18 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
Chris@3
|
19 NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
|
Chris@3
|
20 ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
|
Chris@3
|
21 CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
|
Chris@3
|
22 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
|
Chris@3
|
23 */
|
Chris@3
|
24
|
Chris@31
|
25 #include "CepstralPitchTracker.h"
|
Chris@51
|
26 #include "Cepstrum.h"
|
Chris@47
|
27 #include "MeanFilter.h"
|
Chris@50
|
28 #include "PeakInterpolator.h"
|
Chris@56
|
29 #include "AgentFeeder.h"
|
Chris@3
|
30
|
Chris@26
|
31 #include "vamp-sdk/FFT.h"
|
Chris@26
|
32
|
Chris@3
|
33 #include <vector>
|
Chris@3
|
34 #include <algorithm>
|
Chris@3
|
35
|
Chris@3
|
36 #include <cstdio>
|
Chris@3
|
37 #include <cmath>
|
Chris@3
|
38 #include <complex>
|
Chris@3
|
39
|
Chris@3
|
40 using std::string;
|
Chris@7
|
41 using std::vector;
|
Chris@16
|
42 using Vamp::RealTime;
|
Chris@7
|
43
|
Chris@16
|
44
|
Chris@31
|
45 CepstralPitchTracker::CepstralPitchTracker(float inputSampleRate) :
|
Chris@3
|
46 Plugin(inputSampleRate),
|
Chris@3
|
47 m_channels(0),
|
Chris@3
|
48 m_stepSize(256),
|
Chris@3
|
49 m_blockSize(1024),
|
Chris@3
|
50 m_fmin(50),
|
Chris@25
|
51 m_fmax(900),
|
Chris@18
|
52 m_vflen(1),
|
Chris@3
|
53 m_binFrom(0),
|
Chris@3
|
54 m_binTo(0),
|
Chris@56
|
55 m_bins(0),
|
Chris@56
|
56 m_feeder(0)
|
Chris@3
|
57 {
|
Chris@3
|
58 }
|
Chris@3
|
59
|
Chris@31
|
60 CepstralPitchTracker::~CepstralPitchTracker()
|
Chris@3
|
61 {
|
Chris@56
|
62 delete m_feeder;
|
Chris@3
|
63 }
|
Chris@3
|
64
|
Chris@3
|
65 string
|
Chris@31
|
66 CepstralPitchTracker::getIdentifier() const
|
Chris@3
|
67 {
|
Chris@39
|
68 return "cepstral-pitchtracker";
|
Chris@3
|
69 }
|
Chris@3
|
70
|
Chris@3
|
71 string
|
Chris@31
|
72 CepstralPitchTracker::getName() const
|
Chris@3
|
73 {
|
Chris@39
|
74 return "Cepstral Pitch Tracker";
|
Chris@3
|
75 }
|
Chris@3
|
76
|
Chris@3
|
77 string
|
Chris@31
|
78 CepstralPitchTracker::getDescription() const
|
Chris@3
|
79 {
|
Chris@3
|
80 return "Estimate f0 of monophonic material using a cepstrum method.";
|
Chris@3
|
81 }
|
Chris@3
|
82
|
Chris@3
|
83 string
|
Chris@31
|
84 CepstralPitchTracker::getMaker() const
|
Chris@3
|
85 {
|
Chris@3
|
86 return "Chris Cannam";
|
Chris@3
|
87 }
|
Chris@3
|
88
|
Chris@3
|
89 int
|
Chris@31
|
90 CepstralPitchTracker::getPluginVersion() const
|
Chris@3
|
91 {
|
Chris@3
|
92 // Increment this each time you release a version that behaves
|
Chris@3
|
93 // differently from the previous one
|
Chris@3
|
94 return 1;
|
Chris@3
|
95 }
|
Chris@3
|
96
|
Chris@3
|
97 string
|
Chris@31
|
98 CepstralPitchTracker::getCopyright() const
|
Chris@3
|
99 {
|
Chris@3
|
100 return "Freely redistributable (BSD license)";
|
Chris@3
|
101 }
|
Chris@3
|
102
|
Chris@31
|
103 CepstralPitchTracker::InputDomain
|
Chris@31
|
104 CepstralPitchTracker::getInputDomain() const
|
Chris@3
|
105 {
|
Chris@3
|
106 return FrequencyDomain;
|
Chris@3
|
107 }
|
Chris@3
|
108
|
Chris@3
|
109 size_t
|
Chris@31
|
110 CepstralPitchTracker::getPreferredBlockSize() const
|
Chris@3
|
111 {
|
Chris@3
|
112 return 1024;
|
Chris@3
|
113 }
|
Chris@3
|
114
|
Chris@3
|
115 size_t
|
Chris@31
|
116 CepstralPitchTracker::getPreferredStepSize() const
|
Chris@3
|
117 {
|
Chris@3
|
118 return 256;
|
Chris@3
|
119 }
|
Chris@3
|
120
|
Chris@3
|
121 size_t
|
Chris@31
|
122 CepstralPitchTracker::getMinChannelCount() const
|
Chris@3
|
123 {
|
Chris@3
|
124 return 1;
|
Chris@3
|
125 }
|
Chris@3
|
126
|
Chris@3
|
127 size_t
|
Chris@31
|
128 CepstralPitchTracker::getMaxChannelCount() const
|
Chris@3
|
129 {
|
Chris@3
|
130 return 1;
|
Chris@3
|
131 }
|
Chris@3
|
132
|
Chris@31
|
133 CepstralPitchTracker::ParameterList
|
Chris@31
|
134 CepstralPitchTracker::getParameterDescriptors() const
|
Chris@3
|
135 {
|
Chris@3
|
136 ParameterList list;
|
Chris@3
|
137 return list;
|
Chris@3
|
138 }
|
Chris@3
|
139
|
Chris@3
|
140 float
|
Chris@31
|
141 CepstralPitchTracker::getParameter(string identifier) const
|
Chris@3
|
142 {
|
Chris@3
|
143 return 0.f;
|
Chris@3
|
144 }
|
Chris@3
|
145
|
Chris@3
|
146 void
|
Chris@31
|
147 CepstralPitchTracker::setParameter(string identifier, float value)
|
Chris@3
|
148 {
|
Chris@3
|
149 }
|
Chris@3
|
150
|
Chris@31
|
151 CepstralPitchTracker::ProgramList
|
Chris@31
|
152 CepstralPitchTracker::getPrograms() const
|
Chris@3
|
153 {
|
Chris@3
|
154 ProgramList list;
|
Chris@3
|
155 return list;
|
Chris@3
|
156 }
|
Chris@3
|
157
|
Chris@3
|
158 string
|
Chris@31
|
159 CepstralPitchTracker::getCurrentProgram() const
|
Chris@3
|
160 {
|
Chris@3
|
161 return ""; // no programs
|
Chris@3
|
162 }
|
Chris@3
|
163
|
Chris@3
|
164 void
|
Chris@31
|
165 CepstralPitchTracker::selectProgram(string name)
|
Chris@3
|
166 {
|
Chris@3
|
167 }
|
Chris@3
|
168
|
Chris@31
|
169 CepstralPitchTracker::OutputList
|
Chris@31
|
170 CepstralPitchTracker::getOutputDescriptors() const
|
Chris@3
|
171 {
|
Chris@3
|
172 OutputList outputs;
|
Chris@3
|
173
|
Chris@3
|
174 OutputDescriptor d;
|
Chris@3
|
175
|
Chris@3
|
176 d.identifier = "f0";
|
Chris@3
|
177 d.name = "Estimated f0";
|
Chris@3
|
178 d.description = "Estimated fundamental frequency";
|
Chris@3
|
179 d.unit = "Hz";
|
Chris@3
|
180 d.hasFixedBinCount = true;
|
Chris@3
|
181 d.binCount = 1;
|
Chris@3
|
182 d.hasKnownExtents = true;
|
Chris@3
|
183 d.minValue = m_fmin;
|
Chris@3
|
184 d.maxValue = m_fmax;
|
Chris@3
|
185 d.isQuantized = false;
|
Chris@3
|
186 d.sampleType = OutputDescriptor::FixedSampleRate;
|
Chris@3
|
187 d.sampleRate = (m_inputSampleRate / m_stepSize);
|
Chris@3
|
188 d.hasDuration = false;
|
Chris@3
|
189 outputs.push_back(d);
|
Chris@3
|
190
|
Chris@16
|
191 d.identifier = "notes";
|
Chris@16
|
192 d.name = "Notes";
|
Chris@16
|
193 d.description = "Derived fixed-pitch note frequencies";
|
Chris@16
|
194 d.unit = "Hz";
|
Chris@16
|
195 d.hasFixedBinCount = true;
|
Chris@16
|
196 d.binCount = 1;
|
Chris@16
|
197 d.hasKnownExtents = true;
|
Chris@16
|
198 d.minValue = m_fmin;
|
Chris@16
|
199 d.maxValue = m_fmax;
|
Chris@16
|
200 d.isQuantized = false;
|
Chris@16
|
201 d.sampleType = OutputDescriptor::FixedSampleRate;
|
Chris@16
|
202 d.sampleRate = (m_inputSampleRate / m_stepSize);
|
Chris@16
|
203 d.hasDuration = true;
|
Chris@16
|
204 outputs.push_back(d);
|
Chris@16
|
205
|
Chris@3
|
206 return outputs;
|
Chris@3
|
207 }
|
Chris@3
|
208
|
Chris@3
|
209 bool
|
Chris@31
|
210 CepstralPitchTracker::initialise(size_t channels, size_t stepSize, size_t blockSize)
|
Chris@3
|
211 {
|
Chris@3
|
212 if (channels < getMinChannelCount() ||
|
Chris@3
|
213 channels > getMaxChannelCount()) return false;
|
Chris@3
|
214
|
Chris@31
|
215 // std::cerr << "CepstralPitchTracker::initialise: channels = " << channels
|
Chris@3
|
216 // << ", stepSize = " << stepSize << ", blockSize = " << blockSize
|
Chris@3
|
217 // << std::endl;
|
Chris@3
|
218
|
Chris@3
|
219 m_channels = channels;
|
Chris@3
|
220 m_stepSize = stepSize;
|
Chris@3
|
221 m_blockSize = blockSize;
|
Chris@3
|
222
|
Chris@3
|
223 m_binFrom = int(m_inputSampleRate / m_fmax);
|
Chris@3
|
224 m_binTo = int(m_inputSampleRate / m_fmin);
|
Chris@3
|
225
|
Chris@3
|
226 if (m_binTo >= (int)m_blockSize / 2) {
|
Chris@3
|
227 m_binTo = m_blockSize / 2 - 1;
|
Chris@3
|
228 }
|
Chris@56
|
229 if (m_binFrom >= m_binTo) {
|
Chris@56
|
230 // shouldn't happen except for degenerate samplerate / blocksize combos
|
Chris@56
|
231 m_binFrom = m_binTo - 1;
|
Chris@56
|
232 }
|
Chris@3
|
233
|
Chris@3
|
234 m_bins = (m_binTo - m_binFrom) + 1;
|
Chris@3
|
235
|
Chris@3
|
236 reset();
|
Chris@3
|
237
|
Chris@3
|
238 return true;
|
Chris@3
|
239 }
|
Chris@3
|
240
|
Chris@3
|
241 void
|
Chris@31
|
242 CepstralPitchTracker::reset()
|
Chris@3
|
243 {
|
Chris@56
|
244 delete m_feeder;
|
Chris@56
|
245 m_feeder = new AgentFeeder();
|
Chris@3
|
246 }
|
Chris@3
|
247
|
Chris@3
|
248 void
|
Chris@35
|
249 CepstralPitchTracker::addFeaturesFrom(NoteHypothesis h, FeatureSet &fs)
|
Chris@30
|
250 {
|
Chris@35
|
251 NoteHypothesis::Estimates es = h.getAcceptedEstimates();
|
Chris@30
|
252
|
Chris@35
|
253 for (int i = 0; i < (int)es.size(); ++i) {
|
Chris@30
|
254 Feature f;
|
Chris@30
|
255 f.hasTimestamp = true;
|
Chris@30
|
256 f.timestamp = es[i].time;
|
Chris@30
|
257 f.values.push_back(es[i].freq);
|
Chris@30
|
258 fs[0].push_back(f);
|
Chris@30
|
259 }
|
Chris@30
|
260
|
Chris@30
|
261 Feature nf;
|
Chris@30
|
262 nf.hasTimestamp = true;
|
Chris@30
|
263 nf.hasDuration = true;
|
Chris@35
|
264 NoteHypothesis::Note n = h.getAveragedNote();
|
Chris@30
|
265 nf.timestamp = n.time;
|
Chris@30
|
266 nf.duration = n.duration;
|
Chris@30
|
267 nf.values.push_back(n.freq);
|
Chris@30
|
268 fs[1].push_back(nf);
|
Chris@30
|
269 }
|
Chris@30
|
270
|
Chris@31
|
271 CepstralPitchTracker::FeatureSet
|
Chris@31
|
272 CepstralPitchTracker::process(const float *const *inputBuffers, RealTime timestamp)
|
Chris@3
|
273 {
|
Chris@3
|
274 FeatureSet fs;
|
Chris@3
|
275
|
Chris@51
|
276 double *rawcep = new double[m_blockSize];
|
Chris@51
|
277 double magmean = Cepstrum(m_blockSize).process(inputBuffers[0], rawcep);
|
Chris@3
|
278
|
Chris@3
|
279 int n = m_bins;
|
Chris@3
|
280 double *data = new double[n];
|
Chris@51
|
281 MeanFilter(m_vflen).filterSubsequence
|
Chris@51
|
282 (rawcep, data, m_blockSize, n, m_binFrom);
|
Chris@51
|
283
|
Chris@3
|
284 delete[] rawcep;
|
Chris@3
|
285
|
Chris@3
|
286 double maxval = 0.0;
|
Chris@6
|
287 int maxbin = -1;
|
Chris@3
|
288
|
Chris@3
|
289 for (int i = 0; i < n; ++i) {
|
Chris@3
|
290 if (data[i] > maxval) {
|
Chris@3
|
291 maxval = data[i];
|
Chris@3
|
292 maxbin = i;
|
Chris@3
|
293 }
|
Chris@3
|
294 }
|
Chris@3
|
295
|
Chris@15
|
296 if (maxbin < 0) {
|
Chris@15
|
297 delete[] data;
|
Chris@15
|
298 return fs;
|
Chris@15
|
299 }
|
Chris@15
|
300
|
Chris@15
|
301 double nextPeakVal = 0.0;
|
Chris@15
|
302 for (int i = 1; i+1 < n; ++i) {
|
Chris@15
|
303 if (data[i] > data[i-1] &&
|
Chris@15
|
304 data[i] > data[i+1] &&
|
Chris@15
|
305 i != maxbin &&
|
Chris@15
|
306 data[i] > nextPeakVal) {
|
Chris@15
|
307 nextPeakVal = data[i];
|
Chris@15
|
308 }
|
Chris@15
|
309 }
|
Chris@8
|
310
|
Chris@50
|
311 PeakInterpolator pi;
|
Chris@50
|
312 double cimax = pi.findPeakLocation(data, m_bins, maxbin);
|
Chris@18
|
313 double peakfreq = m_inputSampleRate / (cimax + m_binFrom);
|
Chris@15
|
314
|
Chris@15
|
315 double confidence = 0.0;
|
Chris@51
|
316 double threshold = 0.1; // for magmean
|
Chris@51
|
317
|
Chris@15
|
318 if (nextPeakVal != 0.0) {
|
Chris@27
|
319 confidence = (maxval - nextPeakVal) * 10.0;
|
Chris@25
|
320 if (magmean < threshold) confidence = 0.0;
|
Chris@39
|
321 // std::cerr << "magmean = " << magmean << ", confidence = " << confidence << std::endl;
|
Chris@15
|
322 }
|
Chris@15
|
323
|
Chris@35
|
324 NoteHypothesis::Estimate e;
|
Chris@8
|
325 e.freq = peakfreq;
|
Chris@8
|
326 e.time = timestamp;
|
Chris@15
|
327 e.confidence = confidence;
|
Chris@8
|
328
|
Chris@56
|
329 m_feeder->feed(e);
|
Chris@14
|
330
|
Chris@3
|
331 delete[] data;
|
Chris@3
|
332 return fs;
|
Chris@3
|
333 }
|
Chris@3
|
334
|
Chris@31
|
335 CepstralPitchTracker::FeatureSet
|
Chris@31
|
336 CepstralPitchTracker::getRemainingFeatures()
|
Chris@3
|
337 {
|
Chris@56
|
338 m_feeder->finish();
|
Chris@56
|
339
|
Chris@56
|
340 AgentFeeder::Hypotheses accepted = m_feeder->getAcceptedHypotheses();
|
Chris@56
|
341
|
Chris@3
|
342 FeatureSet fs;
|
Chris@56
|
343 for (int i = 0; i < accepted.size(); ++i) {
|
Chris@56
|
344 addFeaturesFrom(accepted[i], fs);
|
Chris@11
|
345 }
|
Chris@3
|
346 return fs;
|
Chris@3
|
347 }
|