cannam@0:
PercussionOnsetDetector.cpp

00001 /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
cannam@0: 00002 
cannam@0: 00003 /*
cannam@0: 00004     Vamp
cannam@0: 00005 
cannam@0: 00006     An API for audio analysis and feature extraction plugins.
cannam@0: 00007 
cannam@0: 00008     Centre for Digital Music, Queen Mary, University of London.
cannam@0: 00009     Copyright 2006 Chris Cannam.
cannam@0: 00010   
cannam@0: 00011     Permission is hereby granted, free of charge, to any person
cannam@0: 00012     obtaining a copy of this software and associated documentation
cannam@0: 00013     files (the "Software"), to deal in the Software without
cannam@0: 00014     restriction, including without limitation the rights to use, copy,
cannam@0: 00015     modify, merge, publish, distribute, sublicense, and/or sell copies
cannam@0: 00016     of the Software, and to permit persons to whom the Software is
cannam@0: 00017     furnished to do so, subject to the following conditions:
cannam@0: 00018 
cannam@0: 00019     The above copyright notice and this permission notice shall be
cannam@0: 00020     included in all copies or substantial portions of the Software.
cannam@0: 00021 
cannam@0: 00022     THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
cannam@0: 00023     EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
cannam@0: 00024     MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
cannam@0: 00025     NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
cannam@0: 00026     ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
cannam@0: 00027     CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
cannam@0: 00028     WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
cannam@0: 00029 
cannam@0: 00030     Except as contained in this notice, the names of the Centre for
cannam@0: 00031     Digital Music; Queen Mary, University of London; and Chris Cannam
cannam@0: 00032     shall not be used in advertising or otherwise to promote the sale,
cannam@0: 00033     use or other dealings in this Software without prior written
cannam@0: 00034     authorization.
cannam@0: 00035 */
cannam@0: 00036 
cannam@0: 00037 #include "PercussionOnsetDetector.h"
cannam@0: 00038 
cannam@0: 00039 using std::string;
cannam@0: 00040 using std::vector;
cannam@0: 00041 using std::cerr;
cannam@0: 00042 using std::endl;
cannam@0: 00043 
cannam@0: 00044 #include <cmath>
cannam@0: 00045 
cannam@0: 00046 
cannam@0: 00047 PercussionOnsetDetector::PercussionOnsetDetector(float inputSampleRate) :
cannam@0: 00048     Plugin(inputSampleRate),
cannam@0: 00049     m_stepSize(0),
cannam@0: 00050     m_blockSize(0),
cannam@0: 00051     m_threshold(3),
cannam@0: 00052     m_sensitivity(40),
cannam@0: 00053     m_priorMagnitudes(0),
cannam@0: 00054     m_dfMinus1(0),
cannam@0: 00055     m_dfMinus2(0)
cannam@0: 00056 {
cannam@0: 00057 }
cannam@0: 00058 
cannam@0: 00059 PercussionOnsetDetector::~PercussionOnsetDetector()
cannam@0: 00060 {
cannam@0: 00061     delete[] m_priorMagnitudes;
cannam@0: 00062 }
cannam@0: 00063 
cannam@0: 00064 string
cannam@0: 00065 PercussionOnsetDetector::getIdentifier() const
cannam@0: 00066 {
cannam@0: 00067     return "percussiononsets";
cannam@0: 00068 }
cannam@0: 00069 
cannam@0: 00070 string
cannam@0: 00071 PercussionOnsetDetector::getName() const
cannam@0: 00072 {
cannam@0: 00073     return "Simple Percussion Onset Detector";
cannam@0: 00074 }
cannam@0: 00075 
cannam@0: 00076 string
cannam@0: 00077 PercussionOnsetDetector::getDescription() const
cannam@0: 00078 {
cannam@0: 00079     return "Detect percussive note onsets by identifying broadband energy rises";
cannam@0: 00080 }
cannam@0: 00081 
cannam@0: 00082 string
cannam@0: 00083 PercussionOnsetDetector::getMaker() const
cannam@0: 00084 {
cannam@0: 00085     return "Vamp SDK Example Plugins";
cannam@0: 00086 }
cannam@0: 00087 
cannam@0: 00088 int
cannam@0: 00089 PercussionOnsetDetector::getPluginVersion() const
cannam@0: 00090 {
cannam@0: 00091     return 2;
cannam@0: 00092 }
cannam@0: 00093 
cannam@0: 00094 string
cannam@0: 00095 PercussionOnsetDetector::getCopyright() const
cannam@0: 00096 {
cannam@0: 00097     return "Code copyright 2006 Queen Mary, University of London, after Dan Barry et al 2005.  Freely redistributable (BSD license)";
cannam@0: 00098 }
cannam@0: 00099 
cannam@0: 00100 size_t
cannam@0: 00101 PercussionOnsetDetector::getPreferredStepSize() const
cannam@0: 00102 {
cannam@0: 00103     return 0;
cannam@0: 00104 }
cannam@0: 00105 
cannam@0: 00106 size_t
cannam@0: 00107 PercussionOnsetDetector::getPreferredBlockSize() const
cannam@0: 00108 {
cannam@0: 00109     return 1024;
cannam@0: 00110 }
cannam@0: 00111 
cannam@0: 00112 bool
cannam@0: 00113 PercussionOnsetDetector::initialise(size_t channels, size_t stepSize, size_t blockSize)
cannam@0: 00114 {
cannam@0: 00115     if (channels < getMinChannelCount() ||
cannam@0: 00116         channels > getMaxChannelCount()) return false;
cannam@0: 00117 
cannam@0: 00118     m_stepSize = stepSize;
cannam@0: 00119     m_blockSize = blockSize;
cannam@0: 00120 
cannam@0: 00121     m_priorMagnitudes = new float[m_blockSize/2];
cannam@0: 00122 
cannam@0: 00123     for (size_t i = 0; i < m_blockSize/2; ++i) {
cannam@0: 00124         m_priorMagnitudes[i] = 0.f;
cannam@0: 00125     }
cannam@0: 00126 
cannam@0: 00127     m_dfMinus1 = 0.f;
cannam@0: 00128     m_dfMinus2 = 0.f;
cannam@0: 00129 
cannam@0: 00130     return true;
cannam@0: 00131 }
cannam@0: 00132 
cannam@0: 00133 void
cannam@0: 00134 PercussionOnsetDetector::reset()
cannam@0: 00135 {
cannam@0: 00136     for (size_t i = 0; i < m_blockSize/2; ++i) {
cannam@0: 00137         m_priorMagnitudes[i] = 0.f;
cannam@0: 00138     }
cannam@0: 00139 
cannam@0: 00140     m_dfMinus1 = 0.f;
cannam@0: 00141     m_dfMinus2 = 0.f;
cannam@0: 00142 }
cannam@0: 00143 
cannam@0: 00144 PercussionOnsetDetector::ParameterList
cannam@0: 00145 PercussionOnsetDetector::getParameterDescriptors() const
cannam@0: 00146 {
cannam@0: 00147     ParameterList list;
cannam@0: 00148 
cannam@0: 00149     ParameterDescriptor d;
cannam@35: 00150     d.identifier = "threshold";
cannam@35: 00151     d.name = "Energy rise threshold";
cannam@35: 00152     d.description = "Energy rise within a frequency bin necessary to count toward broadband total";
cannam@35: 00153     d.unit = "dB";
cannam@35: 00154     d.minValue = 0;
cannam@35: 00155     d.maxValue = 20;
cannam@35: 00156     d.defaultValue = 3;
cannam@35: 00157     d.isQuantized = false;
cannam@0: 00158     list.push_back(d);
cannam@0: 00159 
cannam@35: 00160     d.identifier = "sensitivity";
cannam@35: 00161     d.name = "Sensitivity";
cannam@35: 00162     d.description = "Sensitivity of peak detector applied to broadband detection function";
cannam@35: 00163     d.unit = "%";
cannam@35: 00164     d.minValue = 0;
cannam@35: 00165     d.maxValue = 100;
cannam@35: 00166     d.defaultValue = 40;
cannam@35: 00167     d.isQuantized = false;
cannam@0: 00168     list.push_back(d);
cannam@0: 00169 
cannam@0: 00170     return list;
cannam@0: 00171 }
cannam@0: 00172 
cannam@0: 00173 float
cannam@0: 00174 PercussionOnsetDetector::getParameter(std::string id) const
cannam@0: 00175 {
cannam@0: 00176     if (id == "threshold") return m_threshold;
cannam@0: 00177     if (id == "sensitivity") return m_sensitivity;
cannam@0: 00178     return 0.f;
cannam@0: 00179 }
cannam@0: 00180 
cannam@0: 00181 void
cannam@0: 00182 PercussionOnsetDetector::setParameter(std::string id, float value)
cannam@0: 00183 {
cannam@0: 00184     if (id == "threshold") {
cannam@0: 00185         if (value < 0) value = 0;
cannam@0: 00186         if (value > 20) value = 20;
cannam@0: 00187         m_threshold = value;
cannam@0: 00188     } else if (id == "sensitivity") {
cannam@0: 00189         if (value < 0) value = 0;
cannam@0: 00190         if (value > 100) value = 100;
cannam@0: 00191         m_sensitivity = value;
cannam@0: 00192     }
cannam@0: 00193 }
cannam@0: 00194 
cannam@0: 00195 PercussionOnsetDetector::OutputList
cannam@0: 00196 PercussionOnsetDetector::getOutputDescriptors() const
cannam@0: 00197 {
cannam@0: 00198     OutputList list;
cannam@0: 00199 
cannam@0: 00200     OutputDescriptor d;
cannam@35: 00201     d.identifier = "onsets";
cannam@35: 00202     d.name = "Onsets";
cannam@35: 00203     d.description = "Percussive note onset locations";
cannam@35: 00204     d.unit = "";
cannam@35: 00205     d.hasFixedBinCount = true;
cannam@35: 00206     d.binCount = 0;
cannam@35: 00207     d.hasKnownExtents = false;
cannam@35: 00208     d.isQuantized = false;
cannam@35: 00209     d.sampleType = OutputDescriptor::VariableSampleRate;
cannam@35: 00210     d.sampleRate = m_inputSampleRate;
cannam@0: 00211     list.push_back(d);
cannam@0: 00212 
cannam@35: 00213     d.identifier = "detectionfunction";
cannam@35: 00214     d.name = "Detection Function";
cannam@35: 00215     d.description = "Broadband energy rise detection function";
cannam@35: 00216     d.binCount = 1;
cannam@35: 00217     d.isQuantized = true;
cannam@35: 00218     d.quantizeStep = 1.0;
cannam@35: 00219     d.sampleType = OutputDescriptor::OneSamplePerStep;
cannam@0: 00220     list.push_back(d);
cannam@0: 00221 
cannam@0: 00222     return list;
cannam@0: 00223 }
cannam@0: 00224 
cannam@0: 00225 PercussionOnsetDetector::FeatureSet
cannam@0: 00226 PercussionOnsetDetector::process(const float *const *inputBuffers,
cannam@0: 00227                                  Vamp::RealTime ts)
cannam@0: 00228 {
cannam@0: 00229     if (m_stepSize == 0) {
cannam@0: 00230         cerr << "ERROR: PercussionOnsetDetector::process: "
cannam@0: 00231              << "PercussionOnsetDetector has not been initialised"
cannam@0: 00232              << endl;
cannam@0: 00233         return FeatureSet();
cannam@0: 00234     }
cannam@0: 00235 
cannam@0: 00236     int count = 0;
cannam@0: 00237 
cannam@0: 00238     for (size_t i = 1; i < m_blockSize/2; ++i) {
cannam@0: 00239 
cannam@0: 00240         float real = inputBuffers[0][i*2];
cannam@0: 00241         float imag = inputBuffers[0][i*2 + 1];
cannam@0: 00242 
cannam@0: 00243         float sqrmag = real * real + imag * imag;
cannam@0: 00244 
cannam@0: 00245         if (m_priorMagnitudes[i] > 0.f) {
cannam@0: 00246             float diff = 10.f * log10f(sqrmag / m_priorMagnitudes[i]);
cannam@0: 00247 
cannam@21: 00248 //        std::cout << "i=" << i << ", sqrmag=" << sqrmag << ", prior=" << m_priorMagnitudes[i] << ", diff=" << diff << ", threshold=" << m_threshold << " " << (diff >= m_threshold ? "[*]" : "") << std::endl;
cannam@0: 00249 
cannam@0: 00250             if (diff >= m_threshold) ++count;
cannam@0: 00251         }
cannam@0: 00252 
cannam@0: 00253         m_priorMagnitudes[i] = sqrmag;
cannam@0: 00254     }
cannam@0: 00255 
cannam@0: 00256     FeatureSet returnFeatures;
cannam@0: 00257 
cannam@0: 00258     Feature detectionFunction;
cannam@35: 00259     detectionFunction.hasTimestamp = false;
cannam@35: 00260     detectionFunction.values.push_back(count);
cannam@0: 00261     returnFeatures[1].push_back(detectionFunction);
cannam@0: 00262 
cannam@0: 00263     if (m_dfMinus2 < m_dfMinus1 &&
cannam@0: 00264         m_dfMinus1 >= count &&
cannam@0: 00265         m_dfMinus1 > ((100 - m_sensitivity) * m_blockSize) / 200) {
cannam@0: 00266 
cannam@21: 00267 //std::cout << "result at " << ts << "! (count == " << count << ", prev == " << m_dfMinus1 << ")" << std::endl;
cannam@21: 00268 
cannam@21: 00269         Feature onset;
cannam@35: 00270         onset.hasTimestamp = true;
cannam@35: 00271         onset.timestamp = ts - Vamp::RealTime::frame2RealTime
cannam@21: 00272             (m_stepSize, int(m_inputSampleRate + 0.5));
cannam@21: 00273         returnFeatures[0].push_back(onset);
cannam@21: 00274     }
cannam@21: 00275 
cannam@21: 00276     m_dfMinus2 = m_dfMinus1;
cannam@21: 00277     m_dfMinus1 = count;
cannam@21: 00278 
cannam@21: 00279     return returnFeatures;
cannam@21: 00280 }
cannam@21: 00281 
cannam@21: 00282 PercussionOnsetDetector::FeatureSet
cannam@21: 00283 PercussionOnsetDetector::getRemainingFeatures()
cannam@21: 00284 {
cannam@21: 00285     return FeatureSet();
cannam@21: 00286 }
cannam@21: 00287 
cannam@0: