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 plugins using Paul Brossier's Aubio library.
|
cannam@0
|
5
|
cannam@0
|
6 Centre for Digital Music, Queen Mary, University of London.
|
cannam@0
|
7 This file copyright 2006 Chris Cannam.
|
cannam@0
|
8
|
cannam@0
|
9 This program is free software; you can redistribute it and/or
|
cannam@0
|
10 modify it under the terms of the GNU General Public License as
|
cannam@0
|
11 published by the Free Software Foundation; either version 2 of the
|
cannam@0
|
12 License, or (at your option) any later version. See the file
|
cannam@0
|
13 COPYING included with this distribution for more information.
|
cannam@0
|
14
|
cannam@0
|
15 */
|
cannam@0
|
16
|
piem@2
|
17 #include <math.h>
|
cannam@0
|
18 #include "Pitch.h"
|
cannam@0
|
19
|
cannam@0
|
20 using std::string;
|
cannam@0
|
21 using std::vector;
|
cannam@0
|
22 using std::cerr;
|
cannam@0
|
23 using std::endl;
|
cannam@0
|
24
|
cannam@23
|
25 static float
|
cannam@23
|
26 getFrequencyForMIDIPitch(int midiPitch)
|
cannam@23
|
27 {
|
cannam@23
|
28 return 440.f * powf(2.0, (float(midiPitch) - 69.0) / 12.0);
|
cannam@23
|
29 }
|
cannam@23
|
30
|
cannam@0
|
31 Pitch::Pitch(float inputSampleRate) :
|
cannam@0
|
32 Plugin(inputSampleRate),
|
cannam@0
|
33 m_ibuf(0),
|
cannam@35
|
34 m_obuf(0),
|
cannam@0
|
35 m_pitchdet(0),
|
cannam@31
|
36 m_pitchtype(PitchYinFFT),
|
cannam@23
|
37 m_minfreq(getFrequencyForMIDIPitch(32)),
|
cannam@23
|
38 m_maxfreq(getFrequencyForMIDIPitch(95)),
|
cannam@23
|
39 m_silence(-90),
|
cannam@23
|
40 m_wrapRange(false),
|
cannam@23
|
41 m_stepSize(0),
|
cannam@32
|
42 m_blockSize(0)
|
cannam@0
|
43 {
|
cannam@0
|
44 }
|
cannam@0
|
45
|
cannam@0
|
46 Pitch::~Pitch()
|
cannam@0
|
47 {
|
cannam@31
|
48 if (m_pitchdet) del_aubio_pitch(m_pitchdet);
|
cannam@0
|
49 if (m_ibuf) del_fvec(m_ibuf);
|
cannam@33
|
50 if (m_obuf) del_fvec(m_obuf);
|
cannam@0
|
51 }
|
cannam@0
|
52
|
cannam@0
|
53 string
|
cannam@13
|
54 Pitch::getIdentifier() const
|
cannam@0
|
55 {
|
cannam@0
|
56 return "aubiopitch";
|
cannam@0
|
57 }
|
cannam@0
|
58
|
cannam@0
|
59 string
|
cannam@13
|
60 Pitch::getName() const
|
cannam@13
|
61 {
|
cannam@13
|
62 return "Aubio Pitch Detector";
|
cannam@13
|
63 }
|
cannam@13
|
64
|
cannam@13
|
65 string
|
cannam@0
|
66 Pitch::getDescription() const
|
cannam@0
|
67 {
|
cannam@13
|
68 return "Track estimated note pitches";
|
cannam@0
|
69 }
|
cannam@0
|
70
|
cannam@0
|
71 string
|
cannam@0
|
72 Pitch::getMaker() const
|
cannam@0
|
73 {
|
cannam@0
|
74 return "Paul Brossier (plugin by Chris Cannam)";
|
cannam@0
|
75 }
|
cannam@0
|
76
|
cannam@0
|
77 int
|
cannam@0
|
78 Pitch::getPluginVersion() const
|
cannam@0
|
79 {
|
cannam@31
|
80 return 3;
|
cannam@0
|
81 }
|
cannam@0
|
82
|
cannam@0
|
83 string
|
cannam@0
|
84 Pitch::getCopyright() const
|
cannam@0
|
85 {
|
cannam@0
|
86 return "GPL";
|
cannam@0
|
87 }
|
cannam@0
|
88
|
cannam@0
|
89 bool
|
cannam@0
|
90 Pitch::initialise(size_t channels, size_t stepSize, size_t blockSize)
|
cannam@0
|
91 {
|
cannam@32
|
92 if (channels != 1) {
|
cannam@32
|
93 std::cerr << "Pitch::initialise: channels must be 1" << std::endl;
|
cannam@32
|
94 return false;
|
cannam@32
|
95 }
|
cannam@32
|
96
|
cannam@0
|
97 m_stepSize = stepSize;
|
cannam@0
|
98 m_blockSize = blockSize;
|
cannam@0
|
99
|
cannam@32
|
100 m_ibuf = new_fvec(stepSize);
|
cannam@33
|
101 m_obuf = new_fvec(1);
|
cannam@0
|
102
|
cannam@37
|
103 reset();
|
cannam@0
|
104
|
cannam@0
|
105 return true;
|
cannam@0
|
106 }
|
cannam@0
|
107
|
cannam@0
|
108 void
|
cannam@0
|
109 Pitch::reset()
|
cannam@0
|
110 {
|
cannam@37
|
111 if (m_pitchdet) del_aubio_pitch(m_pitchdet);
|
cannam@37
|
112
|
cannam@37
|
113 m_pitchdet = new_aubio_pitch
|
cannam@37
|
114 (const_cast<char *>(getAubioNameForPitchType(m_pitchtype)),
|
cannam@37
|
115 m_blockSize,
|
cannam@37
|
116 m_stepSize,
|
cannam@37
|
117 lrintf(m_inputSampleRate));
|
cannam@37
|
118
|
cannam@37
|
119 aubio_pitch_set_unit(m_pitchdet, const_cast<char *>("freq"));
|
cannam@0
|
120 }
|
cannam@0
|
121
|
cannam@0
|
122 size_t
|
cannam@0
|
123 Pitch::getPreferredStepSize() const
|
cannam@0
|
124 {
|
cannam@0
|
125 return 512;
|
cannam@0
|
126 }
|
cannam@0
|
127
|
cannam@0
|
128 size_t
|
cannam@0
|
129 Pitch::getPreferredBlockSize() const
|
cannam@0
|
130 {
|
piem@2
|
131 return 2048;
|
cannam@0
|
132 }
|
cannam@0
|
133
|
cannam@0
|
134 Pitch::ParameterList
|
cannam@0
|
135 Pitch::getParameterDescriptors() const
|
cannam@0
|
136 {
|
cannam@0
|
137 ParameterList list;
|
cannam@0
|
138
|
cannam@0
|
139 ParameterDescriptor desc;
|
cannam@13
|
140 desc.identifier = "pitchtype";
|
cannam@13
|
141 desc.name = "Pitch Detection Function Type";
|
cannam@0
|
142 desc.minValue = 0;
|
cannam@0
|
143 desc.maxValue = 4;
|
cannam@33
|
144 desc.defaultValue = (int)PitchYinFFT;
|
cannam@0
|
145 desc.isQuantized = true;
|
cannam@0
|
146 desc.quantizeStep = 1;
|
cannam@0
|
147 desc.valueNames.push_back("YIN Frequency Estimator");
|
cannam@0
|
148 desc.valueNames.push_back("Spectral Comb");
|
cannam@0
|
149 desc.valueNames.push_back("Schmitt");
|
cannam@0
|
150 desc.valueNames.push_back("Fast Harmonic Comb");
|
cannam@0
|
151 desc.valueNames.push_back("YIN with FFT");
|
cannam@0
|
152 list.push_back(desc);
|
cannam@0
|
153
|
cannam@23
|
154 desc = ParameterDescriptor();
|
cannam@23
|
155 desc.identifier = "minfreq";
|
cannam@23
|
156 desc.name = "Minimum Fundamental Frequency";
|
cannam@23
|
157 desc.minValue = 1;
|
cannam@23
|
158 desc.maxValue = m_inputSampleRate/2;
|
cannam@23
|
159 desc.defaultValue = getFrequencyForMIDIPitch(32);
|
cannam@23
|
160 desc.unit = "Hz";
|
cannam@23
|
161 desc.isQuantized = false;
|
cannam@23
|
162 list.push_back(desc);
|
cannam@23
|
163
|
cannam@23
|
164 desc = ParameterDescriptor();
|
cannam@23
|
165 desc.identifier = "maxfreq";
|
cannam@23
|
166 desc.name = "Maximum Fundamental Frequency";
|
cannam@23
|
167 desc.minValue = 1;
|
cannam@23
|
168 desc.maxValue = m_inputSampleRate/2;
|
cannam@23
|
169 desc.defaultValue = getFrequencyForMIDIPitch(95);
|
cannam@23
|
170 desc.unit = "Hz";
|
cannam@23
|
171 desc.isQuantized = false;
|
cannam@23
|
172 list.push_back(desc);
|
cannam@23
|
173
|
cannam@23
|
174 desc = ParameterDescriptor();
|
cannam@23
|
175 desc.identifier = "wraprange";
|
cannam@23
|
176 desc.name = "Fold Higher or Lower Frequencies into Range";
|
cannam@23
|
177 desc.minValue = 0;
|
cannam@23
|
178 desc.maxValue = 1;
|
cannam@23
|
179 desc.defaultValue = 0;
|
cannam@23
|
180 desc.isQuantized = true;
|
cannam@23
|
181 desc.quantizeStep = 1;
|
cannam@23
|
182 list.push_back(desc);
|
cannam@23
|
183
|
cannam@23
|
184 desc = ParameterDescriptor();
|
cannam@23
|
185 desc.identifier = "silencethreshold";
|
cannam@23
|
186 desc.name = "Silence Threshold";
|
cannam@23
|
187 desc.minValue = -120;
|
cannam@23
|
188 desc.maxValue = 0;
|
cannam@23
|
189 desc.defaultValue = -90;
|
cannam@23
|
190 desc.unit = "dB";
|
cannam@23
|
191 desc.isQuantized = false;
|
cannam@23
|
192 list.push_back(desc);
|
cannam@23
|
193
|
cannam@0
|
194 return list;
|
cannam@0
|
195 }
|
cannam@0
|
196
|
cannam@0
|
197 float
|
cannam@0
|
198 Pitch::getParameter(std::string param) const
|
cannam@0
|
199 {
|
cannam@0
|
200 if (param == "pitchtype") {
|
cannam@0
|
201 return m_pitchtype;
|
cannam@23
|
202 } else if (param == "minfreq") {
|
cannam@23
|
203 return m_minfreq;
|
cannam@23
|
204 } else if (param == "maxfreq") {
|
cannam@23
|
205 return m_maxfreq;
|
cannam@23
|
206 } else if (param == "wraprange") {
|
cannam@23
|
207 return m_wrapRange ? 1.0 : 0.0;
|
cannam@23
|
208 } else if (param == "silencethreshold") {
|
cannam@23
|
209 return m_silence;
|
cannam@0
|
210 } else {
|
cannam@0
|
211 return 0.0;
|
cannam@0
|
212 }
|
cannam@0
|
213 }
|
cannam@0
|
214
|
cannam@0
|
215 void
|
cannam@0
|
216 Pitch::setParameter(std::string param, float value)
|
cannam@0
|
217 {
|
cannam@0
|
218 if (param == "pitchtype") {
|
cannam@0
|
219 switch (lrintf(value)) {
|
cannam@33
|
220 case 0: m_pitchtype = PitchYin; break;
|
cannam@33
|
221 case 1: m_pitchtype = PitchMComb; break;
|
cannam@33
|
222 case 2: m_pitchtype = PitchSchmitt; break;
|
cannam@33
|
223 case 3: m_pitchtype = PitchFComb; break;
|
cannam@33
|
224 case 4: m_pitchtype = PitchYinFFT; break;
|
cannam@0
|
225 }
|
cannam@23
|
226 } else if (param == "minfreq") {
|
cannam@23
|
227 m_minfreq = value;
|
cannam@23
|
228 } else if (param == "maxfreq") {
|
cannam@23
|
229 m_maxfreq = value;
|
cannam@23
|
230 } else if (param == "wraprange") {
|
cannam@23
|
231 m_wrapRange = (value > 0.5);
|
cannam@23
|
232 } else if (param == "silencethreshold") {
|
cannam@23
|
233 m_silence = value;
|
cannam@0
|
234 }
|
cannam@0
|
235 }
|
cannam@0
|
236
|
cannam@0
|
237 Pitch::OutputList
|
cannam@0
|
238 Pitch::getOutputDescriptors() const
|
cannam@0
|
239 {
|
cannam@0
|
240 OutputList list;
|
cannam@0
|
241
|
cannam@0
|
242 OutputDescriptor d;
|
cannam@13
|
243 d.identifier = "frequency";
|
cannam@23
|
244 d.name = "Fundamental Frequency";
|
cannam@0
|
245 d.unit = "Hz";
|
cannam@0
|
246 d.hasFixedBinCount = true;
|
cannam@0
|
247 d.binCount = 1;
|
cannam@0
|
248 d.hasKnownExtents = false;
|
cannam@0
|
249 d.isQuantized = false;
|
cannam@23
|
250 d.sampleType = OutputDescriptor::VariableSampleRate;
|
cannam@23
|
251 d.sampleRate = 0;
|
cannam@23
|
252 if (m_stepSize != 0) {
|
cannam@23
|
253 d.sampleRate = m_inputSampleRate / m_stepSize;
|
cannam@23
|
254 }
|
cannam@0
|
255 list.push_back(d);
|
cannam@0
|
256
|
cannam@0
|
257 return list;
|
cannam@0
|
258 }
|
cannam@0
|
259
|
cannam@0
|
260 Pitch::FeatureSet
|
cannam@12
|
261 Pitch::process(const float *const *inputBuffers,
|
cannam@23
|
262 Vamp::RealTime timestamp)
|
cannam@0
|
263 {
|
cannam@23
|
264 FeatureSet returnFeatures;
|
cannam@23
|
265
|
cannam@23
|
266 if (m_stepSize == 0) {
|
cannam@23
|
267 std::cerr << "Pitch::process: Pitch plugin not initialised" << std::endl;
|
cannam@23
|
268 return returnFeatures;
|
cannam@23
|
269 }
|
cannam@23
|
270
|
cannam@0
|
271 for (size_t i = 0; i < m_stepSize; ++i) {
|
piem@52
|
272 fvec_set_sample(m_ibuf, inputBuffers[0][i], i);
|
cannam@0
|
273 }
|
cannam@0
|
274
|
cannam@33
|
275 aubio_pitch_do(m_pitchdet, m_ibuf, m_obuf);
|
cannam@33
|
276
|
cannam@33
|
277 float freq = m_obuf->data[0];
|
cannam@23
|
278
|
cannam@23
|
279 bool silent = aubio_silence_detection(m_ibuf, m_silence);
|
cannam@23
|
280 if (silent) {
|
cannam@23
|
281 // std::cerr << "(silent)" << std::endl;
|
cannam@23
|
282 return returnFeatures;
|
cannam@23
|
283 }
|
cannam@23
|
284
|
cannam@23
|
285 if (m_wrapRange) {
|
cannam@23
|
286 while (freq > 0 && freq < m_minfreq) {
|
cannam@23
|
287 freq = freq * 2.0;
|
cannam@23
|
288 }
|
cannam@23
|
289 while (freq > m_maxfreq) {
|
cannam@23
|
290 freq = freq / 2.0;
|
cannam@23
|
291 }
|
cannam@23
|
292 }
|
cannam@23
|
293
|
cannam@23
|
294 if (freq < m_minfreq || freq > m_maxfreq) {
|
cannam@23
|
295 return returnFeatures;
|
cannam@23
|
296 }
|
cannam@0
|
297
|
cannam@0
|
298 Feature feature;
|
cannam@23
|
299 feature.hasTimestamp = true;
|
cannam@23
|
300 feature.timestamp = timestamp;
|
cannam@23
|
301 feature.values.push_back(freq);
|
cannam@0
|
302
|
cannam@0
|
303 returnFeatures[0].push_back(feature);
|
cannam@0
|
304 return returnFeatures;
|
cannam@0
|
305 }
|
cannam@0
|
306
|
cannam@0
|
307 Pitch::FeatureSet
|
cannam@0
|
308 Pitch::getRemainingFeatures()
|
cannam@0
|
309 {
|
cannam@0
|
310 return FeatureSet();
|
cannam@0
|
311 }
|
cannam@0
|
312
|