cannam@17: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
cannam@17: 
cannam@17: /*
cannam@17:     Vamp feature extraction plugins using Paul Brossier's Aubio library.
cannam@17: 
cannam@17:     Centre for Digital Music, Queen Mary, University of London.
cannam@17:     This file copyright 2006 Chris Cannam.
cannam@17:     
cannam@17:     This program is free software; you can redistribute it and/or
cannam@17:     modify it under the terms of the GNU General Public License as
cannam@17:     published by the Free Software Foundation; either version 2 of the
cannam@17:     License, or (at your option) any later version.  See the file
cannam@17:     COPYING included with this distribution for more information.
cannam@17: */
cannam@17: 
cannam@17: #include <math.h>
cannam@17: #include "Silence.h"
cannam@17: 
cannam@17: using std::string;
cannam@17: using std::vector;
cannam@17: using std::cerr;
cannam@17: using std::endl;
cannam@17: 
cannam@17: Silence::Silence(float inputSampleRate) :
cannam@17:     Plugin(inputSampleRate),
cannam@17:     m_ibuf(0),
cannam@17:     m_pbuf(0),
cannam@17:     m_tmpptrs(0),
cannam@18:     m_threshold(-80),
cannam@17:     m_prevSilent(false),
cannam@17:     m_first(true)
cannam@17: {
cannam@17: }
cannam@17: 
cannam@17: Silence::~Silence()
cannam@17: {
cannam@17:     if (m_ibuf) del_fvec(m_ibuf);
cannam@17:     if (m_pbuf) del_fvec(m_pbuf);
cannam@17:     if (m_tmpptrs) delete[] m_tmpptrs;
cannam@17: }
cannam@17: 
cannam@17: string
cannam@17: Silence::getIdentifier() const
cannam@17: {
cannam@17:     return "aubiosilence";
cannam@17: }
cannam@17: 
cannam@17: string
cannam@17: Silence::getName() const
cannam@17: {
cannam@17:     return "Aubio Silence Detector";
cannam@17: }
cannam@17: 
cannam@17: string
cannam@17: Silence::getDescription() const
cannam@17: {
cannam@17:     return "Detect levels below a certain threshold";
cannam@17: }
cannam@17: 
cannam@17: string
cannam@17: Silence::getMaker() const
cannam@17: {
cannam@17:     return "Paul Brossier (plugin by Chris Cannam)";
cannam@17: }
cannam@17: 
cannam@17: int
cannam@17: Silence::getPluginVersion() const
cannam@17: {
cannam@17:     return 1;
cannam@17: }
cannam@17: 
cannam@17: string
cannam@17: Silence::getCopyright() const
cannam@17: {
cannam@17:     return "GPL";
cannam@17: }
cannam@17: 
cannam@17: bool
cannam@17: Silence::initialise(size_t channels, size_t stepSize, size_t blockSize)
cannam@17: {
cannam@17:     m_channelCount = channels;
cannam@17:     m_stepSize = stepSize;
cannam@17:     m_blockSize = blockSize;
cannam@17: 
cannam@17:     m_ibuf = new_fvec(stepSize, channels);
cannam@17:     m_pbuf = new_fvec(stepSize, channels);
cannam@17:     m_tmpptrs = new smpl_t *[channels];
cannam@17: 
cannam@17:     return true;
cannam@17: }
cannam@17: 
cannam@17: void
cannam@17: Silence::reset()
cannam@17: {
cannam@17:     m_first = true;
cannam@17: }
cannam@17: 
cannam@17: size_t
cannam@17: Silence::getPreferredStepSize() const
cannam@17: {
cannam@17:     return 1024;
cannam@17: }
cannam@17: 
cannam@17: size_t
cannam@17: Silence::getPreferredBlockSize() const
cannam@17: {
cannam@17:     return 1024;
cannam@17: }
cannam@17: 
cannam@17: Silence::ParameterList
cannam@17: Silence::getParameterDescriptors() const
cannam@17: {
cannam@17:     ParameterList list;
cannam@17:     ParameterDescriptor desc;
cannam@17: 
cannam@17:     desc = ParameterDescriptor();
cannam@17:     desc.identifier = "silencethreshold";
cannam@17:     desc.name = "Silence Threshold";
cannam@17:     desc.minValue = -120;
cannam@17:     desc.maxValue = 0;
cannam@18:     desc.defaultValue = -80;
cannam@17:     desc.unit = "dB";
cannam@17:     desc.isQuantized = false;
cannam@17:     list.push_back(desc);
cannam@17: 
cannam@17:     return list;
cannam@17: }
cannam@17: 
cannam@17: float
cannam@17: Silence::getParameter(std::string param) const
cannam@17: {
cannam@17:     if (param == "silencethreshold") {
cannam@17:         return m_threshold;
cannam@17:     } else {
cannam@17:         return 0.0;
cannam@17:     }
cannam@17: }
cannam@17: 
cannam@17: void
cannam@17: Silence::setParameter(std::string param, float value)
cannam@17: {
cannam@17:     if (param == "silencethreshold") {
cannam@17:         m_threshold = value;
cannam@17:     }
cannam@17: }
cannam@17: 
cannam@17: Silence::OutputList
cannam@17: Silence::getOutputDescriptors() const
cannam@17: {
cannam@17:     OutputList list;
cannam@17: 
cannam@17:     OutputDescriptor d;
cannam@17:     d.identifier = "silencestart";
cannam@18:     d.name = "Beginnings of Silent Regions";
cannam@17:     d.description = "Return a single instant at the point where each silent region begins";
cannam@17:     d.hasFixedBinCount = true;
cannam@17:     d.binCount = 0;
cannam@17:     d.sampleType = OutputDescriptor::VariableSampleRate;
cannam@18:     d.sampleRate = 0;
cannam@17:     list.push_back(d);
cannam@17: 
cannam@17:     d.identifier = "silenceend";
cannam@17:     d.name = "Ends of Silent Regions";
cannam@17:     d.description = "Return a single instant at the point where each silent region ends";
cannam@17:     d.hasFixedBinCount = true;
cannam@17:     d.binCount = 0;
cannam@17:     d.sampleType = OutputDescriptor::VariableSampleRate;
cannam@18:     d.sampleRate = 0;
cannam@17:     list.push_back(d);
cannam@17: 
cannam@17:     d.identifier = "silencelevel";
cannam@17:     d.name = "Silence Test";
cannam@17:     d.description = "Return a function that switches from 1 to 0 when silence falls, and back again when it ends";
cannam@17:     d.hasFixedBinCount = true;
cannam@17:     d.binCount = 1;
cannam@17:     d.hasKnownExtents = true;
cannam@17:     d.minValue = 0;
cannam@17:     d.maxValue = 1;
cannam@17:     d.isQuantized = true;
cannam@17:     d.quantizeStep = 1;
cannam@17:     d.sampleType = OutputDescriptor::VariableSampleRate;
cannam@18:     d.sampleRate = 0;
cannam@17:     list.push_back(d);
cannam@17: 
cannam@17:     return list;
cannam@17: }
cannam@17: 
cannam@17: Silence::FeatureSet
cannam@17: Silence::process(const float *const *inputBuffers,
cannam@17:                  Vamp::RealTime timestamp)
cannam@17: {
cannam@17:     for (size_t i = 0; i < m_stepSize; ++i) {
cannam@17:         for (size_t j = 0; j < m_channelCount; ++j) {
cannam@17:             fvec_write_sample(m_ibuf, inputBuffers[j][i], j, i);
cannam@17:         }
cannam@17:     }
cannam@17: 
cannam@17:     bool silent = aubio_silence_detection(m_ibuf, m_threshold);
cannam@17:     FeatureSet returnFeatures;
cannam@17: 
cannam@17:     if (m_first || m_prevSilent != silent) {
cannam@17: 
cannam@17:         Vamp::RealTime featureStamp = timestamp;
cannam@17: 
cannam@17:         if ((silent && !m_first) || !silent) {
cannam@17:         
cannam@17:             // refine our result
cannam@17: 
cannam@17:             long off = 0;
cannam@17:             size_t incr = 16;
cannam@17:             if (incr > m_stepSize/8) incr = m_stepSize/8;
cannam@17: 
cannam@17:             fvec_t vec;
cannam@17:             vec.length = incr * 4;
cannam@17:             vec.channels = m_channelCount;
cannam@17:             vec.data = m_tmpptrs;
cannam@17:             
cannam@17:             for (size_t i = 0; i < m_stepSize - incr * 4; i += incr) {
cannam@17:                 for (size_t j = 0; j < m_channelCount; ++j) {
cannam@17:                     m_tmpptrs[j] = m_ibuf->data[j] + i;
cannam@17:                 }
cannam@17:                 bool subsilent = aubio_silence_detection(&vec, m_threshold);
cannam@17:                 if (silent == subsilent) {
cannam@17:                     off = i;
cannam@17:                     break;
cannam@17:                 }
cannam@17:             }
cannam@17: 
cannam@17:             if (silent && (off == 0)) {
cannam@17:                 for (size_t i = 0; i < m_stepSize - incr; i += incr) {
cannam@17:                     for (size_t j = 0; j < m_channelCount; ++j) {
cannam@17:                         m_tmpptrs[j] = m_pbuf->data[j] + m_stepSize - i - incr;
cannam@17:                     }
cannam@17:                     bool subsilent = aubio_silence_detection(&vec, m_threshold);
cannam@17:                     if (!subsilent) {
cannam@17:                         off = -(long)i;
cannam@17:                         break;
cannam@17:                     }
cannam@17:                 }
cannam@17:             } else {
cannam@17:             }                
cannam@17: 
cannam@17:             featureStamp = timestamp + Vamp::RealTime::frame2RealTime
cannam@17:                 (off, lrintf(m_inputSampleRate));
cannam@17:         }
cannam@17: 
cannam@17:         Feature feature;
cannam@17:         feature.hasTimestamp = true;
cannam@17:         feature.timestamp = featureStamp;
cannam@17:         feature.values.push_back(silent ? 0 : 1);
cannam@17:         returnFeatures[2].push_back(feature);
cannam@17:         feature.values.clear();
cannam@17: 
cannam@17:         if (silent) {
cannam@17:             returnFeatures[0].push_back(feature);
cannam@17:         } else {
cannam@17:             returnFeatures[1].push_back(feature);
cannam@17:         }            
cannam@17: 
cannam@17:         m_prevSilent = silent;
cannam@17:         m_first = false;
cannam@17:     }
cannam@17: 
cannam@17:     // swap ibuf and pbuf data pointers, so that this block's data is
cannam@17:     // available in pbuf when processing the next block, without
cannam@17:     // having to allocate new storage for it
cannam@17:     smpl_t **tmpdata = m_ibuf->data;
cannam@17:     m_ibuf->data = m_pbuf->data;
cannam@17:     m_pbuf->data = tmpdata;
cannam@17: 
cannam@17:     return returnFeatures;
cannam@17: }
cannam@17: 
cannam@17: Silence::FeatureSet
cannam@17: Silence::getRemainingFeatures()
cannam@17: {
cannam@17:     return FeatureSet();
cannam@17: }
cannam@17: