|
cannam@6
|
1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
|
|
cannam@6
|
2 <html>
|
|
cannam@6
|
3 <head>
|
|
cannam@6
|
4 <link rel="stylesheet" media="screen" type="text/css" href="/screen.css"/>
|
|
cannam@6
|
5 <link rel="icon" type="image/png" href="/images/waveform.png"/>
|
|
cannam@6
|
6 <link rel="shortcut" type="image/png" href="/images/waveform.png"/>
|
|
cannam@6
|
7 <title>Vamp Example Plugins: User Documentation</title>
|
|
cannam@6
|
8 <meta name="robots" content="index"/>
|
|
cannam@6
|
9 </head>
|
|
cannam@6
|
10 <body>
|
|
cannam@6
|
11 <h1 id="header"><span>Vamp Example Plugins</span></h1>
|
|
cannam@6
|
12
|
|
cannam@6
|
13 <p>The <code>vamp-example-plugins</code> library contains a number of
|
|
cannam@6
|
14 <a href="http://www.vamp-plugins.org/">Vamp audio analysis
|
|
cannam@6
|
15 plugins</a> provided as part of the Vamp plugin SDK.
|
|
cannam@6
|
16
|
|
cannam@6
|
17 </p>
|
|
cannam@6
|
18 <p>These are simple, but sometimes useful, plugins whose source code you
|
|
cannam@6
|
19 are free to study and reuse in any proprietary or non-proprietary
|
|
cannam@6
|
20 plugins of your own without any licensing obligation.
|
|
cannam@6
|
21 </p>
|
|
cannam@6
|
22 <p>User documentation for the individual plugins in this library follows.
|
|
cannam@6
|
23 </p>
|
|
cannam@6
|
24 <div class="toc2">1. <a href="#amplitudefollower">Amplitude Follower</a></div>
|
|
cannam@6
|
25 <div class="toc2">2. <a href="#fixedtempo">Simple Fixed Tempo Estimator</a></div>
|
|
cannam@6
|
26 <div class="toc2">3. <a href="#percussiononsets">Simple Percussion Onset Detector</a></div>
|
|
cannam@6
|
27 <div class="toc2">4. <a href="#powerspectrum">Simple Power Spectrum</a></div>
|
|
cannam@6
|
28 <div class="toc2">5. <a href="#spectralcentroid">Spectral Centroid</a></div>
|
|
cannam@6
|
29 <div class="toc2">6. <a href="#zerocrossing">Zero Crossings</a></div>
|
|
cannam@6
|
30
|
|
cannam@6
|
31 <div class="oddcontent"><a name="amplitudefollower"></a><h2>1. Amplitude Follower</h2>
|
|
cannam@6
|
32
|
|
cannam@6
|
33 <p><b>System identifier</b> – <code>vamp-example-plugins:amplitudefollower</code><br>
|
|
cannam@6
|
34 <b>RDF URI</b> – <a href="http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#amplitudefollower">http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#amplitudefollower</a>
|
|
cannam@6
|
35 </p>
|
|
cannam@6
|
36 <p>Amplitude Follower tracks and returns the amplitude of the audio
|
|
cannam@6
|
37 signal sample by sample, returning peak values block by block.
|
|
cannam@6
|
38 </p>
|
|
cannam@6
|
39 </div><div class="evencontent"><a name="toc2"></a><h3>1.1. Parameters</h3>
|
|
cannam@6
|
40
|
|
cannam@6
|
41 <p><b>Attack time</b> (seconds) – The 60dB convergence time for an increase in amplitude.<br>
|
|
cannam@6
|
42 <b>Release time</b> (seconds) – The 60dB convergence time for a decrease in amplitude.
|
|
cannam@6
|
43 </p>
|
|
cannam@6
|
44 <p>For example, if you feed the plugin with a simple step function that
|
|
cannam@6
|
45 jumps from level A to level B, then the output will start off as A,
|
|
cannam@6
|
46 then at the moment of stepping it will start to converge exponentially
|
|
cannam@6
|
47 to B, reaching with 60dB of the actual value within the time specified
|
|
cannam@6
|
48 by the Attack time parameter.
|
|
cannam@6
|
49 </p>
|
|
cannam@6
|
50 <p>Similarly, if the plugin's input then steps down from B to A, the
|
|
cannam@6
|
51 output will start converging at the moment of stepping, reaching
|
|
cannam@6
|
52 within 60dB of the new value within the time specified by the Release
|
|
cannam@6
|
53 time parameter.
|
|
cannam@6
|
54 </p>
|
|
cannam@6
|
55 </div><div class="oddcontent"><a name="toc3"></a><h3>1.2. Outputs</h3>
|
|
cannam@6
|
56
|
|
cannam@6
|
57 </div><div class="evencontent"><a name="toc4"></a><h4>1.2.1. Amplitude</h4>
|
|
cannam@6
|
58
|
|
cannam@6
|
59 <p>The peak tracked amplitude (in volts) for the current processing block.
|
|
cannam@6
|
60 </p>
|
|
cannam@6
|
61 </div><div class="oddcontent"><a name="toc5"></a><h3>1.3. References and Credits</h3>
|
|
cannam@6
|
62
|
|
cannam@6
|
63 <p>Amplitude Follower uses a method from the SuperCollider audio
|
|
cannam@6
|
64 processing language. It was implemented as a Vamp plugin by Dan
|
|
cannam@6
|
65 Stowell.
|
|
cannam@6
|
66 </p>
|
|
cannam@6
|
67 </div><div class="evencontent"><a name="fixedtempo"></a><h2>2. Simple Fixed Tempo Estimator</h2>
|
|
cannam@6
|
68
|
|
cannam@6
|
69 <p><b>System identifier</b> – <code>vamp-example-plugins:fixedtempo</code><br>
|
|
cannam@6
|
70 <b>RDF URI</b> – <a href="http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#fixedtempo">http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#fixedtempo</a>
|
|
cannam@6
|
71 </p>
|
|
cannam@6
|
72 <p>Simple Fixed Tempo Estimator analyses a fragment of audio and
|
|
cannam@6
|
73 estimates its tempo. It assumes that its input is of fixed tempo, and
|
|
cannam@6
|
74 it analyses only the first (small but configurable number of) seconds
|
|
cannam@6
|
75 before returning a result, discarding all subsequent input.
|
|
cannam@6
|
76 </p>
|
|
cannam@6
|
77 <p>The plugin calculates an overall energy rise function across a series
|
|
cannam@6
|
78 of short frequency-domain input frames, takes the autocorrelation of
|
|
cannam@6
|
79 this function, filters it to stress possible metrical patterns,
|
|
cannam@6
|
80 locates peaks, and converts from autocorrelation lag to the
|
|
cannam@6
|
81 corresponding tempo.
|
|
cannam@6
|
82 </p>
|
|
cannam@6
|
83 <p>The filtering process involves searching for peaks at simple
|
|
cannam@6
|
84 metrically related intervals (at a given autocorrelation lag as well
|
|
cannam@6
|
85 as at 0.5, 2, and 4 times that lag), boosting each peak that shows
|
|
cannam@6
|
86 strong related peaks. A simplistic perceptual curve is also applied
|
|
cannam@6
|
87 in order to increase the probability of detecting a "likely" tempo.
|
|
cannam@6
|
88 For improved tempo precision, each tempo with strong related peaks is
|
|
cannam@6
|
89 averaged with the tempi calculated from those peaks.
|
|
cannam@6
|
90 </p>
|
|
cannam@6
|
91 <p>The method is best suited for 4/4 pop and dance rhythms.
|
|
cannam@6
|
92 </p>
|
|
cannam@6
|
93 <p>This plugin returns many of its intermediate calculations as
|
|
cannam@6
|
94 additional outputs, as well as the most favoured tempo. Although as a
|
|
cannam@6
|
95 tempo estimator it's still fairly primitive, it is intended to provide
|
|
cannam@6
|
96 a useful example of a slightly more complex feature extraction plugin
|
|
cannam@6
|
97 than the other examples, as well as one that returns several different
|
|
cannam@6
|
98 types of output at a time.
|
|
cannam@6
|
99 </p>
|
|
cannam@6
|
100 </div><div class="oddcontent"><a name="toc7"></a><h3>2.1. Parameters</h3>
|
|
cannam@6
|
101
|
|
cannam@6
|
102 <p><b>Minimum estimated tempo</b>, <b>Maximum estimated tempo</b> (bpm) – These
|
|
cannam@6
|
103 parameters control the range of values within which the tempo
|
|
cannam@6
|
104 estimator will return its estimate.
|
|
cannam@6
|
105 </p>
|
|
cannam@6
|
106 <p><b>Input duration to study</b> (seconds) – The tempo estimator uses only the
|
|
cannam@6
|
107 first part of its input, discarding any that follows. This parameter
|
|
cannam@6
|
108 controls how much input it will use. There is no value in increasing
|
|
cannam@6
|
109 this beyond 8x the duration of the slowest returned beat. The default
|
|
cannam@6
|
110 of 10 seconds is likely to be appropriate for most purposes.
|
|
cannam@6
|
111 </p>
|
|
cannam@6
|
112 </div><div class="evencontent"><a name="toc8"></a><h3>2.2. Outputs</h3>
|
|
cannam@6
|
113
|
|
cannam@6
|
114 </div><div class="oddcontent"><a name="toc9"></a><h4>2.2.1. Tempo</h4>
|
|
cannam@6
|
115
|
|
cannam@6
|
116 <p>The tempo estimator's best guess at the tempo of its input, in beats
|
|
cannam@6
|
117 per minute.
|
|
cannam@6
|
118 </p>
|
|
cannam@6
|
119 <p>This is returned as a feature whose timestamp and duration cover the
|
|
cannam@6
|
120 range of the input which was used in estimating the tempo, with a
|
|
cannam@6
|
121 single value containing the tempo.
|
|
cannam@6
|
122 </p>
|
|
cannam@6
|
123 </div><div class="evencontent"><a name="toc10"></a><h4>2.2.2. Tempo candidates</h4>
|
|
cannam@6
|
124
|
|
cannam@6
|
125 <p>Several guesses at the possible tempo. This output is returned as a
|
|
cannam@6
|
126 single feature whose timestamp and duration cover the range of the
|
|
cannam@6
|
127 input which was used in estimating the tempo, with up to 10 bins
|
|
cannam@6
|
128 containing one tempo value in each bin, with the "best guess" tempo in
|
|
cannam@6
|
129 bin 0.
|
|
cannam@6
|
130 </p>
|
|
cannam@6
|
131 </div><div class="oddcontent"><a name="toc11"></a><h4>2.2.3. Detection function</h4>
|
|
cannam@6
|
132
|
|
cannam@6
|
133 <p>The basic onset detection function used in tempo estimation.
|
|
cannam@6
|
134 </p>
|
|
cannam@6
|
135 </div><div class="evencontent"><a name="toc12"></a><h4>2.2.4. Autocorrelation function</h4>
|
|
cannam@6
|
136
|
|
cannam@6
|
137 <p>The autocorrelation of the onset detection function.
|
|
cannam@6
|
138 </p>
|
|
cannam@6
|
139 </div><div class="oddcontent"><a name="toc13"></a><h4>2.2.5. Filtered Autocorrelation</h4>
|
|
cannam@6
|
140
|
|
cannam@6
|
141 <p>The autocorrelation after filtering to boost values with possible
|
|
cannam@6
|
142 metrically related peaks and to apply perceptual weighting. The peak
|
|
cannam@6
|
143 value of this function is the one that will be used as the "best
|
|
cannam@6
|
144 guess".
|
|
cannam@6
|
145 </p>
|
|
cannam@6
|
146 </div><div class="evencontent"><a name="toc14"></a><h3>2.3. References and Credits</h3>
|
|
cannam@6
|
147
|
|
cannam@6
|
148 <p>Simple Fixed Tempo Estimator uses a method derived from work by
|
|
cannam@6
|
149 Matthew Davies: see for example M. E. P. Davies and M. D. Plumbley,
|
|
cannam@6
|
150 <i>Beat Tracking With A Two State Model</i>, in Proceedings of the IEEE
|
|
cannam@6
|
151 International Conference on Acoustics, Speech and Signal Processing
|
|
cannam@6
|
152 2005. This plugin, made by Chris Cannam, is only an unsubtle
|
|
cannam@6
|
153 simplification of a small part of the published method.
|
|
cannam@6
|
154 </p>
|
|
cannam@6
|
155 <p>The Queen Mary plugin set
|
|
cannam@6
|
156 (<a href="http://www.elec.qmul.ac.uk/digitalmusic/downloads/index.html#qm-vamp-plugins">http://www.elec.qmul.ac.uk/digitalmusic/downloads/index.html#qm-vamp-plugins</a>)
|
|
cannam@6
|
157 contains a Tempo and Beat Tracker plugin by Matthew Davies providing a
|
|
cannam@6
|
158 more realistic implementation.
|
|
cannam@6
|
159 </p>
|
|
cannam@6
|
160 </div><div class="oddcontent"><a name="percussiononsets"></a><h2>3. Simple Percussion Onset Detector</h2>
|
|
cannam@6
|
161
|
|
cannam@6
|
162 <p><b>System identifier</b> – <code>vamp-example-plugins:percussiononsets</code><br>
|
|
cannam@6
|
163 <b>RDF URI</b> – <a href="http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#percussiononsets">http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#percussiononsets</a>
|
|
cannam@6
|
164 </p>
|
|
cannam@6
|
165 <p>Simple Percussion Onset Detector estimates the locations of percussive
|
|
cannam@6
|
166 onsets in the audio signal.
|
|
cannam@6
|
167 </p>
|
|
cannam@6
|
168 <p>The principle is to exploit the broadband nature of noisy percussive
|
|
cannam@6
|
169 onsets by identifying only those frames in which the energy rise shows
|
|
cannam@6
|
170 a broadband profile.
|
|
cannam@6
|
171 </p>
|
|
cannam@6
|
172 <p>The plugin takes a series of frequency domain frames, and examines
|
|
cannam@6
|
173 each frame to count the number of bins whose energy content has
|
|
cannam@6
|
174 increased by more than a certain threshold since the prior frame.
|
|
cannam@6
|
175 Frames in which this number is at a peak relative to prior and
|
|
cannam@6
|
176 following frames and also exceeds another threshold value are
|
|
cannam@6
|
177 classified as percussive onsets.
|
|
cannam@6
|
178 </p>
|
|
cannam@6
|
179 </div><div class="evencontent"><a name="toc16"></a><h3>3.1. Parameters</h3>
|
|
cannam@6
|
180
|
|
cannam@6
|
181 <p><b>Energy rise threshold</b> (dB) – The rise in energy within a bin from one
|
|
cannam@6
|
182 frame to the next that is required for a bin to be counted toward the
|
|
cannam@6
|
183 detection function's bin count. This roughly corresponds to how
|
|
cannam@6
|
184 "loud" a percussive sound must be in order to be detected.
|
|
cannam@6
|
185 </p>
|
|
cannam@6
|
186 <p><b>Sensitivity</b> (%) – The proportion of bins that must exceed the energy
|
|
cannam@6
|
187 rise threshold in order for an onset to be detected (at frames in
|
|
cannam@6
|
188 which the detection function peaks). This roughly corresponds to how
|
|
cannam@6
|
189 "noisy" a percussive sound must be in order to be detected.
|
|
cannam@6
|
190 </p>
|
|
cannam@6
|
191 </div><div class="oddcontent"><a name="toc17"></a><h3>3.2. Outputs</h3>
|
|
cannam@6
|
192
|
|
cannam@6
|
193 </div><div class="evencontent"><a name="toc18"></a><h4>3.2.1. Onsets</h4>
|
|
cannam@6
|
194
|
|
cannam@6
|
195 <p>The estimated onset locations.
|
|
cannam@6
|
196 </p>
|
|
cannam@6
|
197 </div><div class="oddcontent"><a name="toc19"></a><h4>3.2.2. Detection Function</h4>
|
|
cannam@6
|
198
|
|
cannam@6
|
199 <p>The energy rise detection function whose peaks were used to estimate
|
|
cannam@6
|
200 onset locations.
|
|
cannam@6
|
201 </p>
|
|
cannam@6
|
202 </div><div class="evencontent"><a name="toc20"></a><h3>3.3. References and Credits</h3>
|
|
cannam@6
|
203
|
|
cannam@6
|
204 <p>The method used in Simple Percussion Onset Detector was described in
|
|
cannam@6
|
205 Dan Barry, Derry Fitzgerald, Eugene Coyle and
|
|
cannam@6
|
206 Bob Lawlor, <i>Drum Source Separation using Percussive Feature Detection and
|
|
cannam@6
|
207 Spectral Modulation</i>, ISSC 2005. The plugin was made by Chris Cannam.
|
|
cannam@6
|
208 </p>
|
|
cannam@6
|
209 </div><div class="oddcontent"><a name="powerspectrum"></a><h2>4. Simple Power Spectrum</h2>
|
|
cannam@6
|
210
|
|
cannam@6
|
211 <p><b>System identifier</b> – <code>vamp-example-plugins:powerspectrum</code><br>
|
|
cannam@6
|
212 <b>RDF URI</b> – <a href="http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#powerspectrum">http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#powerspectrum</a>
|
|
cannam@6
|
213 </p>
|
|
cannam@6
|
214 <p>Simple Power Spectrum returns a power spectrum calculated from
|
|
cannam@6
|
215 windowed short-time Fourier transforms of the input audio. (The power
|
|
cannam@6
|
216 spectrum for a frame consists of a sequence of the squares of the
|
|
cannam@6
|
217 magnitudes of the complex values for each frequency bin in the result
|
|
cannam@6
|
218 of the Fourier transform.)
|
|
cannam@6
|
219 </p>
|
|
cannam@6
|
220 <p>This very simple plugin is an illustration of the fact that if a
|
|
cannam@6
|
221 plugin requests frequency-domain input, its input will already be in
|
|
cannam@6
|
222 the form needed for a spectrum such as this. The plugin has no work
|
|
cannam@6
|
223 left to do except to calculate the squared magnitude from the
|
|
cannam@6
|
224 cartesian complex representation.
|
|
cannam@6
|
225 </p>
|
|
cannam@6
|
226 <p>This plugin also illustrates how to return "grid-type" visualisation
|
|
cannam@6
|
227 data from a Vamp plugin.
|
|
cannam@6
|
228 </p>
|
|
cannam@6
|
229 </div><div class="evencontent"><a name="toc22"></a><h3>4.1. Parameters</h3>
|
|
cannam@6
|
230
|
|
cannam@6
|
231 <p>None.
|
|
cannam@6
|
232 </p>
|
|
cannam@6
|
233 </div><div class="oddcontent"><a name="toc23"></a><h3>4.2. Outputs</h3>
|
|
cannam@6
|
234
|
|
cannam@6
|
235 </div><div class="evencontent"><a name="toc24"></a><h4>4.2.1. Power Spectrum</h4>
|
|
cannam@6
|
236
|
|
cannam@6
|
237 <p>The power spectrum calculated from the input frame. This output
|
|
cannam@6
|
238 returns a single feature per processing block, containing
|
|
cannam@6
|
239 blocksize/2+1 power values corresponding to the FFT bins from DC to
|
|
cannam@6
|
240 Nyquist inclusive. The DC bin is always returned.
|
|
cannam@6
|
241 </p>
|
|
cannam@6
|
242 </div><div class="oddcontent"><a name="spectralcentroid"></a><h2>5. Spectral Centroid</h2>
|
|
cannam@6
|
243
|
|
cannam@6
|
244 <p><b>System identifier</b> – <code>vamp-example-plugins:spectralcentroid</code><br>
|
|
cannam@6
|
245 <b>RDF URI</b> – <a href="http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#spectralcentroid">http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#spectralcentroid</a>
|
|
cannam@6
|
246 </p>
|
|
cannam@6
|
247 <p>Spectral Centroid calculates the "centre of gravity" of the frequency
|
|
cannam@6
|
248 spectrum for each input frame.
|
|
cannam@6
|
249 </p>
|
|
cannam@6
|
250 </div><div class="evencontent"><a name="toc26"></a><h3>5.1. Parameters</h3>
|
|
cannam@6
|
251
|
|
cannam@6
|
252 <p>None.
|
|
cannam@6
|
253 </p>
|
|
cannam@6
|
254 </div><div class="oddcontent"><a name="toc27"></a><h3>5.2. Outputs</h3>
|
|
cannam@6
|
255
|
|
cannam@6
|
256 </div><div class="evencontent"><a name="toc28"></a><h4>5.2.1. Log Frequency Centroid</h4>
|
|
cannam@6
|
257
|
|
cannam@6
|
258 <p>The centroid of the log-weighted frequency spectrum. That is, the sum
|
|
cannam@6
|
259 across Fourier transform output bins of the logarithm of the bin
|
|
cannam@6
|
260 frequency multiplied by the bin magnitude, divided by the sum of the
|
|
cannam@6
|
261 bin magnitudes, and the inverse logarithm taken so as to give the
|
|
cannam@6
|
262 result as a frequency in Hz.
|
|
cannam@6
|
263 </p>
|
|
cannam@6
|
264 </div><div class="oddcontent"><a name="toc29"></a><h4>5.2.2. Linear Frequency Centroid</h4>
|
|
cannam@6
|
265
|
|
cannam@6
|
266 <p>The centroid of the linear-weighted frequency spectrum. That is, the
|
|
cannam@6
|
267 sum across Fourier transform output bins of the bin frequency
|
|
cannam@6
|
268 multiplied by the bin magnitude, divided by the sum of the bin
|
|
cannam@6
|
269 magnitudes. The result is a frequency in Hz.
|
|
cannam@6
|
270 </p>
|
|
cannam@6
|
271 </div><div class="evencontent"><a name="zerocrossing"></a><h2>6. Zero Crossings</h2>
|
|
cannam@6
|
272
|
|
cannam@6
|
273 <p><b>System identifier</b> – <code>vamp-example-plugins:zerocrossing</code><br>
|
|
cannam@6
|
274 <b>RDF URI</b> – <a href="http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#zerocrossing">http://vamp-plugins.org/rdf/plugins/vamp-example-plugins#zerocrossing</a>
|
|
cannam@6
|
275 </p>
|
|
cannam@6
|
276 <p>Zero Crossings calculates the positions and density of "zero-crossing"
|
|
cannam@6
|
277 points in an audio waveform. For the purposes of this plugin, that
|
|
cannam@6
|
278 means those positions at which the sampled value switches from
|
|
cannam@6
|
279 zero-or-less to greater-than-zero, or vice versa.
|
|
cannam@6
|
280 </p>
|
|
cannam@6
|
281 </div><div class="oddcontent"><a name="toc31"></a><h3>6.1. Parameters</h3>
|
|
cannam@6
|
282
|
|
cannam@6
|
283 <p>None.
|
|
cannam@6
|
284 </p>
|
|
cannam@6
|
285 </div><div class="evencontent"><a name="toc32"></a><h3>6.2. Outputs</h3>
|
|
cannam@6
|
286
|
|
cannam@6
|
287 </div><div class="oddcontent"><a name="toc33"></a><h4>6.2.1. Zero Crossing Counts</h4>
|
|
cannam@6
|
288
|
|
cannam@6
|
289 <p>The number of zero-crossing points found in the current block of
|
|
cannam@6
|
290 samples, as a single-valued feature returned per processing block.
|
|
cannam@6
|
291 </p>
|
|
cannam@6
|
292 </div><div class="evencontent"><a name="toc34"></a><h4>6.2.2. Zero Crossings</h4>
|
|
cannam@6
|
293
|
|
cannam@6
|
294 <p>The locations of zero-crossing points, returning one feature
|
|
cannam@6
|
295 timestamped to the zero-crossing location, without values, for each
|
|
cannam@6
|
296 crossing point.
|
|
cannam@6
|
297 </p>
|
|
cannam@6
|
298 <p></p>
|
|
cannam@6
|
299
|
|
cannam@6
|
300 </div>
|
|
cannam@6
|
301 </body>
|
|
cannam@6
|
302 </html>
|
