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