1 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */
6 An API for audio analysis and feature extraction plugins.
8 Centre for Digital Music, Queen Mary, University of London.
9 Copyright 2006 Chris Cannam.
11 Permission is hereby granted, free of charge, to any person
12 obtaining a copy of this software and associated documentation
13 files (the "Software"), to deal in the Software without
14 restriction, including without limitation the rights to use, copy,
15 modify, merge, publish, distribute, sublicense, and/or sell copies
16 of the Software, and to permit persons to whom the Software is
17 furnished to do so, subject to the following conditions:
19 The above copyright notice and this permission notice shall be
20 included in all copies or substantial portions of the Software.
22 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
23 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
24 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
25 NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
26 ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
27 CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
28 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
30 Except as contained in this notice, the names of the Centre for
31 Digital Music; Queen Mary, University of London; and Chris Cannam
32 shall not be used in advertising or otherwise to promote the sale,
33 use or other dealings in this Software without prior written
37 #ifndef _VAMP_SDK_PLUGIN_H_
38 #define _VAMP_SDK_PLUGIN_H_
44 #include "PluginBase.h"
47 #include "plugguard.h"
48 _VAMP_SDK_PLUGSPACE_BEGIN(Plugin.h)
53 * \class Plugin Plugin.h <vamp-sdk/Plugin.h>
55 * Vamp::Plugin is a base class for plugin instance classes
56 * that provide feature extraction from audio or related data.
58 * In most cases, the input will be audio and the output will be a
59 * stream of derived data at a lower sampling resolution than the
62 * Note that this class inherits several abstract methods from
63 * PluginBase. These must be implemented by the subclass.
68 * Feature extraction plugins are managed differently from real-time
69 * plugins (such as VST effects). The main difference is that the
70 * parameters for a feature extraction plugin are configured before
71 * the plugin is used, and do not change during use.
73 * 1. Host constructs the plugin, passing it the input sample rate.
74 * The plugin may do basic initialisation, but should not do anything
75 * computationally expensive at this point. You must make sure your
76 * plugin is cheap to construct, otherwise you'll seriously affect the
77 * startup performance of almost all hosts. If you have serious
78 * initialisation to do, the proper place is in initialise() (step 5).
80 * 2. Host may query the plugin's available outputs.
82 * 3. Host queries programs and parameter descriptors, and may set
83 * some or all of them. Parameters that are not explicitly set should
84 * take their default values as specified in the parameter descriptor.
85 * When a program is set, the parameter values may change and the host
86 * will re-query them to check.
88 * 4. Host queries the preferred step size, block size and number of
89 * channels. These may all vary depending on the parameter values.
90 * (Note however that you cannot make the number of distinct outputs
91 * dependent on parameter values.)
93 * 5. Plugin is properly initialised with a call to initialise. This
94 * fixes the step size, block size, and number of channels, as well as
95 * all of the parameter and program settings. If the values passed in
96 * to initialise do not match the plugin's advertised preferred values
97 * from step 4, the plugin may refuse to initialise and return false
98 * (although if possible it should accept the new values). Any
99 * computationally expensive setup code should take place here.
101 * 6. Host finally checks the number of values, resolution, extents
102 * etc per output (which may vary depending on the number of channels,
103 * step size and block size as well as the parameter values).
105 * 7. Host will repeatedly call the process method to pass in blocks
106 * of input data. This method may return features extracted from that
107 * data (if the plugin is causal).
109 * 8. Host will call getRemainingFeatures exactly once, after all the
110 * input data has been processed. This may return any non-causal or
113 * 9. At any point after initialise was called, the host may
114 * optionally call the reset method and restart processing. (This
115 * does not mean it can change the parameters, which are fixed from
116 * initialise until destruction.)
118 * A plugin does not need to handle the case where setParameter or
119 * selectProgram is called after initialise has been called. It's the
120 * host's responsibility not to do that. Similarly, the plugin may
121 * safely assume that initialise is called no more than once.
124 class Plugin : public PluginBase
127 virtual ~Plugin() { }
130 * Initialise a plugin to prepare it for use with the given number
131 * of input channels, step size (window increment, in sample
132 * frames) and block size (window size, in sample frames).
134 * The input sample rate should have been already specified at
137 * Return true for successful initialisation, false if the number
138 * of input channels, step size and/or block size cannot be
141 virtual bool initialise(size_t inputChannels,
143 size_t blockSize) = 0;
146 * Reset the plugin after use, to prepare it for another clean
147 * run. Not called for the first initialisation (i.e. initialise
148 * must also do a reset).
150 virtual void reset() = 0;
152 enum InputDomain { TimeDomain, FrequencyDomain };
155 * Get the plugin's required input domain. If this is TimeDomain,
156 * the samples provided to the process() function (below) will be
157 * in the time domain, as for a traditional audio processing
158 * plugin. If this is FrequencyDomain, the host will carry out a
159 * windowed FFT of size equal to the negotiated block size on the
160 * data before passing the frequency bin data in to process().
161 * The plugin does not get to choose the window type -- the host
162 * will either let the user do so, or will use a Hanning window.
164 virtual InputDomain getInputDomain() const = 0;
167 * Get the preferred block size (window size -- the number of
168 * sample frames passed in each block to the process() function).
169 * This should be called before initialise().
171 * A plugin that can handle any block size may return 0. The
172 * final block size will be set in the initialise() call.
174 virtual size_t getPreferredBlockSize() const { return 0; }
177 * Get the preferred step size (window increment -- the distance
178 * in sample frames between the start frames of consecutive blocks
179 * passed to the process() function) for the plugin. This should
180 * be called before initialise().
182 * A plugin may return 0 if it has no particular interest in the
183 * step size. In this case, the host should make the step size
184 * equal to the block size if the plugin is accepting input in the
185 * time domain. If the plugin is accepting input in the frequency
186 * domain, the host may use any step size. The final step size
187 * will be set in the initialise() call.
189 virtual size_t getPreferredStepSize() const { return 0; }
192 * Get the minimum supported number of input channels.
194 virtual size_t getMinChannelCount() const { return 1; }
197 * Get the maximum supported number of input channels.
199 virtual size_t getMaxChannelCount() const { return 1; }
201 struct OutputDescriptor
204 * The name of the output, in computer-usable form. Should be
205 * reasonably short and without whitespace or punctuation, using
206 * the characters [a-zA-Z0-9_-] only.
207 * Example: "zero_crossing_count"
209 std::string identifier;
212 * The human-readable name of the output.
213 * Example: "Zero Crossing Counts"
218 * A human-readable short text describing the output. May be
219 * empty if the name has said it all already.
220 * Example: "The number of zero crossing points per processing block"
222 std::string description;
225 * The unit of the output, in human-readable form.
230 * True if the output has the same number of values per sample
231 * for every output sample. Outputs for which this is false
232 * are unlikely to be very useful in a general-purpose host.
234 bool hasFixedBinCount;
237 * The number of values per result of the output. Undefined
238 * if hasFixedBinCount is false. If this is zero, the output
239 * is point data (i.e. only the time of each output is of
240 * interest, the value list will be empty).
245 * The (human-readable) names of each of the bins, if
246 * appropriate. This is always optional.
248 std::vector<std::string> binNames;
251 * True if the results in each output bin fall within a fixed
252 * numeric range (minimum and maximum values). Undefined if
255 bool hasKnownExtents;
258 * Minimum value of the results in the output. Undefined if
259 * hasKnownExtents is false or binCount is zero.
264 * Maximum value of the results in the output. Undefined if
265 * hasKnownExtents is false or binCount is zero.
270 * True if the output values are quantized to a particular
271 * resolution. Undefined if binCount is zero.
276 * Quantization resolution of the output values (e.g. 1.0 if
277 * they are all integers). Undefined if isQuantized is false
278 * or binCount is zero.
284 /// Results from each process() align with that call's block start
287 /// Results are evenly spaced in time (sampleRate specified below)
290 /// Results are unevenly spaced and have individual timestamps
295 * Positioning in time of the output results.
297 SampleType sampleType;
300 * Sample rate of the output results, as samples per second.
301 * Undefined if sampleType is OneSamplePerStep.
303 * If sampleType is VariableSampleRate and this value is
304 * non-zero, then it may be used to calculate a resolution for
305 * the output (i.e. the "duration" of each sample, in time,
306 * will be 1/sampleRate seconds). It's recommended to set
307 * this to zero if that behaviour is not desired.
312 * True if the returned results for this output are known to
313 * have a duration field.
317 OutputDescriptor() : // defaults for mandatory non-class-type members
318 hasFixedBinCount(false), hasKnownExtents(false), isQuantized(false),
319 sampleType(OneSamplePerStep), hasDuration(false) { }
322 typedef std::vector<OutputDescriptor> OutputList;
325 * Get the outputs of this plugin. An output's index in this list
326 * is used as its numeric index when looking it up in the
327 * FeatureSet returned from the process() call.
329 virtual OutputList getOutputDescriptors() const = 0;
334 * True if an output feature has its own timestamp. This is
335 * mandatory if the output has VariableSampleRate, optional if
336 * the output has FixedSampleRate, and unused if the output
337 * has OneSamplePerStep.
342 * Timestamp of the output feature. This is mandatory if the
343 * output has VariableSampleRate or if the output has
344 * FixedSampleRate and hasTimestamp is true, and unused
350 * True if an output feature has a specified duration. This
351 * is optional if the output has VariableSampleRate or
352 * FixedSampleRate, and and unused if the output has
358 * Duration of the output feature. This is mandatory if the
359 * output has VariableSampleRate or FixedSampleRate and
360 * hasDuration is true, and unused otherwise.
365 * Results for a single sample of this feature. If the output
366 * hasFixedBinCount, there must be the same number of values
367 * as the output's binCount count.
369 std::vector<float> values;
372 * Label for the sample of this feature.
376 Feature() : // defaults for mandatory non-class-type members
377 hasTimestamp(false), hasDuration(false) { }
380 typedef std::vector<Feature> FeatureList;
382 typedef std::map<int, FeatureList> FeatureSet; // key is output no
385 * Process a single block of input data.
387 * If the plugin's inputDomain is TimeDomain, inputBuffers will
388 * point to one array of floats per input channel, and each of
389 * these arrays will contain blockSize consecutive audio samples
390 * (the host will zero-pad as necessary). The timestamp in this
391 * case will be the real time in seconds of the start of the
392 * supplied block of samples.
394 * If the plugin's inputDomain is FrequencyDomain, inputBuffers
395 * will point to one array of floats per input channel, and each
396 * of these arrays will contain blockSize/2+1 consecutive pairs of
397 * real and imaginary component floats corresponding to bins
398 * 0..(blockSize/2) of the FFT output. That is, bin 0 (the first
399 * pair of floats) contains the DC output, up to bin blockSize/2
400 * which contains the Nyquist-frequency output. There will
401 * therefore be blockSize+2 floats per channel in total. The
402 * timestamp will be the real time in seconds of the centre of the
403 * FFT input window (i.e. the very first block passed to process
404 * might contain the FFT of half a block of zero samples and the
405 * first half-block of the actual data, with a timestamp of zero).
407 * Return any features that have become available after this
408 * process call. (These do not necessarily have to fall within
409 * the process block, except for OneSamplePerStep outputs.)
411 virtual FeatureSet process(const float *const *inputBuffers,
412 RealTime timestamp) = 0;
415 * After all blocks have been processed, calculate and return any
416 * remaining features derived from the complete input.
418 virtual FeatureSet getRemainingFeatures() = 0;
421 * Used to distinguish between Vamp::Plugin and other potential
422 * sibling subclasses of PluginBase. Do not reimplement this
423 * function in your subclass.
425 virtual std::string getType() const { return "Feature Extraction Plugin"; }
428 Plugin(float inputSampleRate) :
429 m_inputSampleRate(inputSampleRate) { }
431 float m_inputSampleRate;
436 _VAMP_SDK_PLUGSPACE_END(Plugin.h)