cannam@17
|
1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
|
cannam@17
|
2
|
cannam@17
|
3 /*
|
cannam@17
|
4 Vamp feature extraction plugins using Paul Brossier's Aubio library.
|
cannam@17
|
5
|
cannam@17
|
6 Centre for Digital Music, Queen Mary, University of London.
|
cannam@20
|
7 This file copyright 2006-2008 Chris Cannam and QMUL.
|
cannam@17
|
8
|
cannam@17
|
9 This program is free software; you can redistribute it and/or
|
cannam@17
|
10 modify it under the terms of the GNU General Public License as
|
cannam@17
|
11 published by the Free Software Foundation; either version 2 of the
|
cannam@17
|
12 License, or (at your option) any later version. See the file
|
cannam@17
|
13 COPYING included with this distribution for more information.
|
cannam@17
|
14 */
|
cannam@17
|
15
|
cannam@17
|
16 #include <math.h>
|
cannam@17
|
17 #include "Silence.h"
|
cannam@17
|
18
|
cannam@17
|
19 using std::string;
|
cannam@17
|
20 using std::vector;
|
cannam@17
|
21 using std::cerr;
|
cannam@17
|
22 using std::endl;
|
cannam@17
|
23
|
cannam@20
|
24 Silence::Silence(float inputSampleRate, unsigned int apiVersion) :
|
cannam@17
|
25 Plugin(inputSampleRate),
|
cannam@20
|
26 m_apiVersion(apiVersion),
|
cannam@17
|
27 m_ibuf(0),
|
cannam@17
|
28 m_pbuf(0),
|
cannam@17
|
29 m_tmpptrs(0),
|
cannam@18
|
30 m_threshold(-80),
|
cannam@17
|
31 m_prevSilent(false),
|
cannam@17
|
32 m_first(true)
|
cannam@17
|
33 {
|
cannam@20
|
34 if (m_apiVersion == 1) {
|
cannam@20
|
35 cerr << "vamp-aubio: WARNING: using compatibility version 1 of the Vamp API for silence\n"
|
cannam@20
|
36 << "detector plugin: upgrade your host to v2 for proper duration support" << endl;
|
cannam@20
|
37 }
|
cannam@17
|
38 }
|
cannam@17
|
39
|
cannam@17
|
40 Silence::~Silence()
|
cannam@17
|
41 {
|
cannam@17
|
42 if (m_ibuf) del_fvec(m_ibuf);
|
cannam@17
|
43 if (m_pbuf) del_fvec(m_pbuf);
|
cannam@17
|
44 if (m_tmpptrs) delete[] m_tmpptrs;
|
cannam@17
|
45 }
|
cannam@17
|
46
|
cannam@17
|
47 string
|
cannam@17
|
48 Silence::getIdentifier() const
|
cannam@17
|
49 {
|
cannam@17
|
50 return "aubiosilence";
|
cannam@17
|
51 }
|
cannam@17
|
52
|
cannam@17
|
53 string
|
cannam@17
|
54 Silence::getName() const
|
cannam@17
|
55 {
|
cannam@17
|
56 return "Aubio Silence Detector";
|
cannam@17
|
57 }
|
cannam@17
|
58
|
cannam@17
|
59 string
|
cannam@17
|
60 Silence::getDescription() const
|
cannam@17
|
61 {
|
cannam@17
|
62 return "Detect levels below a certain threshold";
|
cannam@17
|
63 }
|
cannam@17
|
64
|
cannam@17
|
65 string
|
cannam@17
|
66 Silence::getMaker() const
|
cannam@17
|
67 {
|
cannam@17
|
68 return "Paul Brossier (plugin by Chris Cannam)";
|
cannam@17
|
69 }
|
cannam@17
|
70
|
cannam@17
|
71 int
|
cannam@17
|
72 Silence::getPluginVersion() const
|
cannam@17
|
73 {
|
cannam@20
|
74 if (m_apiVersion == 1) return 2;
|
cannam@20
|
75 return 3;
|
cannam@17
|
76 }
|
cannam@17
|
77
|
cannam@17
|
78 string
|
cannam@17
|
79 Silence::getCopyright() const
|
cannam@17
|
80 {
|
cannam@17
|
81 return "GPL";
|
cannam@17
|
82 }
|
cannam@17
|
83
|
cannam@17
|
84 bool
|
cannam@17
|
85 Silence::initialise(size_t channels, size_t stepSize, size_t blockSize)
|
cannam@17
|
86 {
|
cannam@17
|
87 m_channelCount = channels;
|
cannam@17
|
88 m_stepSize = stepSize;
|
cannam@17
|
89 m_blockSize = blockSize;
|
cannam@17
|
90
|
cannam@17
|
91 m_ibuf = new_fvec(stepSize, channels);
|
cannam@17
|
92 m_pbuf = new_fvec(stepSize, channels);
|
cannam@17
|
93 m_tmpptrs = new smpl_t *[channels];
|
cannam@17
|
94
|
cannam@17
|
95 return true;
|
cannam@17
|
96 }
|
cannam@17
|
97
|
cannam@17
|
98 void
|
cannam@17
|
99 Silence::reset()
|
cannam@17
|
100 {
|
cannam@17
|
101 m_first = true;
|
cannam@17
|
102 }
|
cannam@17
|
103
|
cannam@17
|
104 size_t
|
cannam@17
|
105 Silence::getPreferredStepSize() const
|
cannam@17
|
106 {
|
cannam@17
|
107 return 1024;
|
cannam@17
|
108 }
|
cannam@17
|
109
|
cannam@17
|
110 size_t
|
cannam@17
|
111 Silence::getPreferredBlockSize() const
|
cannam@17
|
112 {
|
cannam@17
|
113 return 1024;
|
cannam@17
|
114 }
|
cannam@17
|
115
|
cannam@17
|
116 Silence::ParameterList
|
cannam@17
|
117 Silence::getParameterDescriptors() const
|
cannam@17
|
118 {
|
cannam@17
|
119 ParameterList list;
|
cannam@17
|
120 ParameterDescriptor desc;
|
cannam@17
|
121
|
cannam@17
|
122 desc = ParameterDescriptor();
|
cannam@17
|
123 desc.identifier = "silencethreshold";
|
cannam@17
|
124 desc.name = "Silence Threshold";
|
cannam@17
|
125 desc.minValue = -120;
|
cannam@17
|
126 desc.maxValue = 0;
|
cannam@18
|
127 desc.defaultValue = -80;
|
cannam@17
|
128 desc.unit = "dB";
|
cannam@17
|
129 desc.isQuantized = false;
|
cannam@17
|
130 list.push_back(desc);
|
cannam@17
|
131
|
cannam@17
|
132 return list;
|
cannam@17
|
133 }
|
cannam@17
|
134
|
cannam@17
|
135 float
|
cannam@17
|
136 Silence::getParameter(std::string param) const
|
cannam@17
|
137 {
|
cannam@17
|
138 if (param == "silencethreshold") {
|
cannam@17
|
139 return m_threshold;
|
cannam@17
|
140 } else {
|
cannam@17
|
141 return 0.0;
|
cannam@17
|
142 }
|
cannam@17
|
143 }
|
cannam@17
|
144
|
cannam@17
|
145 void
|
cannam@17
|
146 Silence::setParameter(std::string param, float value)
|
cannam@17
|
147 {
|
cannam@17
|
148 if (param == "silencethreshold") {
|
cannam@17
|
149 m_threshold = value;
|
cannam@17
|
150 }
|
cannam@17
|
151 }
|
cannam@17
|
152
|
cannam@17
|
153 Silence::OutputList
|
cannam@17
|
154 Silence::getOutputDescriptors() const
|
cannam@17
|
155 {
|
cannam@17
|
156 OutputList list;
|
cannam@17
|
157
|
cannam@17
|
158 OutputDescriptor d;
|
cannam@17
|
159
|
cannam@20
|
160 if (m_apiVersion == 1) {
|
cannam@20
|
161
|
cannam@20
|
162 d.identifier = "silencestart";
|
cannam@20
|
163 d.name = "Beginnings of Silent Regions";
|
cannam@20
|
164 d.description = "Return a single instant at the point where each silent region begins";
|
cannam@20
|
165 d.hasFixedBinCount = true;
|
cannam@20
|
166 d.binCount = 0;
|
cannam@20
|
167 d.hasKnownExtents = false;
|
cannam@20
|
168 d.sampleType = OutputDescriptor::VariableSampleRate;
|
cannam@20
|
169 d.sampleRate = 0;
|
cannam@20
|
170 list.push_back(d);
|
cannam@20
|
171
|
cannam@20
|
172 d.identifier = "silenceend";
|
cannam@20
|
173 d.name = "Ends of Silent Regions";
|
cannam@20
|
174 d.description = "Return a single instant at the point where each silent region ends";
|
cannam@20
|
175 d.hasFixedBinCount = true;
|
cannam@20
|
176 d.binCount = 0;
|
cannam@20
|
177 d.hasKnownExtents = false;
|
cannam@20
|
178 d.sampleType = OutputDescriptor::VariableSampleRate;
|
cannam@20
|
179 d.sampleRate = 0;
|
cannam@20
|
180 list.push_back(d);
|
cannam@20
|
181
|
cannam@20
|
182 } else {
|
cannam@20
|
183
|
cannam@20
|
184 d.identifier = "silent";
|
cannam@20
|
185 d.name = "Silent Regions";
|
cannam@20
|
186 d.description = "Return an interval covering each silent region";
|
cannam@20
|
187 d.hasFixedBinCount = true;
|
cannam@20
|
188 d.binCount = 0;
|
cannam@20
|
189 d.hasKnownExtents = false;
|
cannam@20
|
190 d.sampleType = OutputDescriptor::VariableSampleRate;
|
cannam@20
|
191 d.sampleRate = 0;
|
cannam@22
|
192 d.hasDuration = true;
|
cannam@20
|
193 list.push_back(d);
|
cannam@20
|
194
|
cannam@20
|
195 d.identifier = "noisy";
|
cannam@20
|
196 d.name = "Non-Silent Regions";
|
cannam@20
|
197 d.description = "Return an interval covering each non-silent region";
|
cannam@20
|
198 d.hasFixedBinCount = true;
|
cannam@20
|
199 d.binCount = 0;
|
cannam@20
|
200 d.hasKnownExtents = false;
|
cannam@20
|
201 d.sampleType = OutputDescriptor::VariableSampleRate;
|
cannam@20
|
202 d.sampleRate = 0;
|
cannam@22
|
203 d.hasDuration = true;
|
cannam@20
|
204 list.push_back(d);
|
cannam@20
|
205 }
|
cannam@17
|
206
|
cannam@17
|
207 d.identifier = "silencelevel";
|
cannam@17
|
208 d.name = "Silence Test";
|
cannam@17
|
209 d.description = "Return a function that switches from 1 to 0 when silence falls, and back again when it ends";
|
cannam@17
|
210 d.hasFixedBinCount = true;
|
cannam@17
|
211 d.binCount = 1;
|
cannam@17
|
212 d.hasKnownExtents = true;
|
cannam@17
|
213 d.minValue = 0;
|
cannam@17
|
214 d.maxValue = 1;
|
cannam@17
|
215 d.isQuantized = true;
|
cannam@17
|
216 d.quantizeStep = 1;
|
cannam@17
|
217 d.sampleType = OutputDescriptor::VariableSampleRate;
|
cannam@18
|
218 d.sampleRate = 0;
|
cannam@17
|
219 list.push_back(d);
|
cannam@17
|
220
|
cannam@17
|
221 return list;
|
cannam@17
|
222 }
|
cannam@17
|
223
|
cannam@17
|
224 Silence::FeatureSet
|
cannam@17
|
225 Silence::process(const float *const *inputBuffers,
|
cannam@17
|
226 Vamp::RealTime timestamp)
|
cannam@17
|
227 {
|
cannam@17
|
228 for (size_t i = 0; i < m_stepSize; ++i) {
|
cannam@17
|
229 for (size_t j = 0; j < m_channelCount; ++j) {
|
cannam@17
|
230 fvec_write_sample(m_ibuf, inputBuffers[j][i], j, i);
|
cannam@17
|
231 }
|
cannam@17
|
232 }
|
cannam@17
|
233
|
cannam@17
|
234 bool silent = aubio_silence_detection(m_ibuf, m_threshold);
|
cannam@17
|
235 FeatureSet returnFeatures;
|
cannam@17
|
236
|
cannam@17
|
237 if (m_first || m_prevSilent != silent) {
|
cannam@17
|
238
|
cannam@17
|
239 Vamp::RealTime featureStamp = timestamp;
|
cannam@17
|
240
|
cannam@17
|
241 if ((silent && !m_first) || !silent) {
|
cannam@17
|
242
|
cannam@17
|
243 // refine our result
|
cannam@17
|
244
|
cannam@17
|
245 long off = 0;
|
cannam@17
|
246 size_t incr = 16;
|
cannam@17
|
247 if (incr > m_stepSize/8) incr = m_stepSize/8;
|
cannam@17
|
248
|
cannam@17
|
249 fvec_t vec;
|
cannam@17
|
250 vec.length = incr * 4;
|
cannam@17
|
251 vec.channels = m_channelCount;
|
cannam@17
|
252 vec.data = m_tmpptrs;
|
cannam@17
|
253
|
cannam@17
|
254 for (size_t i = 0; i < m_stepSize - incr * 4; i += incr) {
|
cannam@17
|
255 for (size_t j = 0; j < m_channelCount; ++j) {
|
cannam@17
|
256 m_tmpptrs[j] = m_ibuf->data[j] + i;
|
cannam@17
|
257 }
|
cannam@17
|
258 bool subsilent = aubio_silence_detection(&vec, m_threshold);
|
cannam@17
|
259 if (silent == subsilent) {
|
cannam@17
|
260 off = i;
|
cannam@17
|
261 break;
|
cannam@17
|
262 }
|
cannam@17
|
263 }
|
cannam@17
|
264
|
cannam@17
|
265 if (silent && (off == 0)) {
|
cannam@17
|
266 for (size_t i = 0; i < m_stepSize - incr; i += incr) {
|
cannam@17
|
267 for (size_t j = 0; j < m_channelCount; ++j) {
|
cannam@17
|
268 m_tmpptrs[j] = m_pbuf->data[j] + m_stepSize - i - incr;
|
cannam@17
|
269 }
|
cannam@17
|
270 bool subsilent = aubio_silence_detection(&vec, m_threshold);
|
cannam@17
|
271 if (!subsilent) {
|
cannam@17
|
272 off = -(long)i;
|
cannam@17
|
273 break;
|
cannam@17
|
274 }
|
cannam@17
|
275 }
|
cannam@17
|
276 } else {
|
cannam@17
|
277 }
|
cannam@17
|
278
|
cannam@17
|
279 featureStamp = timestamp + Vamp::RealTime::frame2RealTime
|
cannam@17
|
280 (off, lrintf(m_inputSampleRate));
|
cannam@17
|
281 }
|
cannam@17
|
282
|
cannam@17
|
283 Feature feature;
|
cannam@17
|
284 feature.hasTimestamp = true;
|
cannam@17
|
285 feature.timestamp = featureStamp;
|
cannam@17
|
286 feature.values.push_back(silent ? 0 : 1);
|
cannam@17
|
287 returnFeatures[2].push_back(feature);
|
cannam@20
|
288
|
cannam@17
|
289 feature.values.clear();
|
cannam@17
|
290
|
cannam@20
|
291 if (m_apiVersion == 1) {
|
cannam@20
|
292 if (silent) {
|
cannam@20
|
293 returnFeatures[0].push_back(feature);
|
cannam@20
|
294 } else {
|
cannam@20
|
295 returnFeatures[1].push_back(feature);
|
cannam@20
|
296 }
|
cannam@17
|
297 } else {
|
cannam@20
|
298 if (!m_first) {
|
cannam@20
|
299 feature.timestamp = m_lastChange;
|
cannam@20
|
300 feature.hasDuration = true;
|
cannam@20
|
301 feature.duration = featureStamp - m_lastChange;
|
cannam@20
|
302 if (silent) {
|
cannam@20
|
303 // becoming silent, so this is a non-silent region
|
cannam@20
|
304 returnFeatures[1].push_back(feature);
|
cannam@20
|
305 } else {
|
cannam@20
|
306 // becoming non-silent, so this is a silent region
|
cannam@20
|
307 returnFeatures[0].push_back(feature);
|
cannam@20
|
308 }
|
cannam@20
|
309 }
|
cannam@20
|
310 m_lastChange = featureStamp;
|
cannam@20
|
311 }
|
cannam@17
|
312
|
cannam@17
|
313 m_prevSilent = silent;
|
cannam@17
|
314 m_first = false;
|
cannam@17
|
315 }
|
cannam@17
|
316
|
cannam@17
|
317 // swap ibuf and pbuf data pointers, so that this block's data is
|
cannam@17
|
318 // available in pbuf when processing the next block, without
|
cannam@17
|
319 // having to allocate new storage for it
|
cannam@17
|
320 smpl_t **tmpdata = m_ibuf->data;
|
cannam@17
|
321 m_ibuf->data = m_pbuf->data;
|
cannam@17
|
322 m_pbuf->data = tmpdata;
|
cannam@17
|
323
|
cannam@20
|
324 m_lastTimestamp = timestamp;
|
cannam@20
|
325
|
cannam@17
|
326 return returnFeatures;
|
cannam@17
|
327 }
|
cannam@17
|
328
|
cannam@17
|
329 Silence::FeatureSet
|
cannam@17
|
330 Silence::getRemainingFeatures()
|
cannam@17
|
331 {
|
cannam@20
|
332 FeatureSet returnFeatures;
|
cannam@20
|
333
|
cannam@20
|
334 if (m_prevSilent) {
|
cannam@20
|
335 if (m_lastTimestamp > m_lastChange) {
|
cannam@20
|
336 Feature feature;
|
cannam@20
|
337 feature.hasTimestamp = true;
|
cannam@20
|
338 feature.timestamp = m_lastChange;
|
cannam@20
|
339 feature.hasDuration = true;
|
cannam@20
|
340 feature.duration = m_lastTimestamp - m_lastChange;
|
cannam@20
|
341 if (m_prevSilent) {
|
cannam@20
|
342 returnFeatures[0].push_back(feature);
|
cannam@20
|
343 } else {
|
cannam@20
|
344 returnFeatures[1].push_back(feature);
|
cannam@20
|
345 }
|
cannam@20
|
346 }
|
cannam@20
|
347 }
|
cannam@20
|
348
|
cannam@17
|
349 return FeatureSet();
|
cannam@17
|
350 }
|
cannam@17
|
351
|