ardour/livetrax
13
0
Fork 0
Code

NO-OP: remove trailing whitespace

Browse Source
This commit is contained in:
Robin Gareus 2019-09-03 04:52:01 +02:00
parent b340dc7282
commit 58ee66e924
Signed by: rgareus
GPG Key ID: A090BCE02CF57F04
25 changed files with 208 additions and 210 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

56
libs/vamp-pyin/LocalCandidatePYIN.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -106,7 +106,7 @@ LocalCandidatePYIN::getPreferredBlockSize() const
return 2048; return 2048;
} }
size_t size_t
LocalCandidatePYIN::getPreferredStepSize() const LocalCandidatePYIN::getPreferredStepSize() const
{ {
return 256; return 256;
@ -128,7 +128,7 @@ LocalCandidatePYIN::ParameterList
LocalCandidatePYIN::getParameterDescriptors() const LocalCandidatePYIN::getParameterDescriptors() const
{ {
ParameterList list; ParameterList list;
ParameterDescriptor d; ParameterDescriptor d;
d.identifier = "threshdistr"; d.identifier = "threshdistr";
@ -196,7 +196,7 @@ LocalCandidatePYIN::getParameter(string identifier) const
} }
void void
LocalCandidatePYIN::setParameter(string identifier, float value) LocalCandidatePYIN::setParameter(string identifier, float value)
{ {
if (identifier == "threshdistr") if (identifier == "threshdistr")
{ {
@ -268,7 +268,7 @@ LocalCandidatePYIN::initialise(size_t channels, size_t stepSize, size_t blockSiz
m_channels = channels; m_channels = channels;
m_stepSize = stepSize; m_stepSize = stepSize;
m_blockSize = blockSize; m_blockSize = blockSize;
reset(); reset();
return true; return true;
@ -276,10 +276,10 @@ LocalCandidatePYIN::initialise(size_t channels, size_t stepSize, size_t blockSiz
void void
LocalCandidatePYIN::reset() LocalCandidatePYIN::reset()
{ {
m_pitchProb.clear(); m_pitchProb.clear();
m_timestamp.clear(); m_timestamp.clear();
/* /*
std::cerr << "LocalCandidatePYIN::reset" std::cerr << "LocalCandidatePYIN::reset"
<< ", blockSize = " << m_blockSize << ", blockSize = " << m_blockSize
<< std::endl; << std::endl;
@ -291,25 +291,25 @@ LocalCandidatePYIN::process(const float *const *inputBuffers, RealTime timestamp
{ {
int offset = m_preciseTime == 1.0 ? m_blockSize/2 : m_blockSize/4; int offset = m_preciseTime == 1.0 ? m_blockSize/2 : m_blockSize/4;
timestamp = timestamp + Vamp::RealTime::frame2RealTime(offset, lrintf(m_inputSampleRate)); timestamp = timestamp + Vamp::RealTime::frame2RealTime(offset, lrintf(m_inputSampleRate));
double *dInputBuffers = new double[m_blockSize]; double *dInputBuffers = new double[m_blockSize];
for (size_t i = 0; i < m_blockSize; ++i) dInputBuffers[i] = inputBuffers[0][i]; for (size_t i = 0; i < m_blockSize; ++i) dInputBuffers[i] = inputBuffers[0][i];
size_t yinBufferSize = m_blockSize/2; size_t yinBufferSize = m_blockSize/2;
double* yinBuffer = new double[yinBufferSize]; double* yinBuffer = new double[yinBufferSize];
if (!m_preciseTime) YinUtil::fastDifference(dInputBuffers, yinBuffer, yinBufferSize); if (!m_preciseTime) YinUtil::fastDifference(dInputBuffers, yinBuffer, yinBufferSize);
else YinUtil::slowDifference(dInputBuffers, yinBuffer, yinBufferSize); else YinUtil::slowDifference(dInputBuffers, yinBuffer, yinBufferSize);
delete [] dInputBuffers; delete [] dInputBuffers;
YinUtil::cumulativeDifference(yinBuffer, yinBufferSize); YinUtil::cumulativeDifference(yinBuffer, yinBufferSize);
float minFrequency = 60; float minFrequency = 60;
float maxFrequency = 900; float maxFrequency = 900;
vector<double> peakProbability = YinUtil::yinProb(yinBuffer, vector<double> peakProbability = YinUtil::yinProb(yinBuffer,
m_threshDistr, m_threshDistr,
yinBufferSize, yinBufferSize,
m_inputSampleRate/maxFrequency, m_inputSampleRate/maxFrequency,
m_inputSampleRate/minFrequency); m_inputSampleRate/minFrequency);
vector<pair<double, double> > tempPitchProb; vector<pair<double, double> > tempPitchProb;
@ -317,7 +317,7 @@ LocalCandidatePYIN::process(const float *const *inputBuffers, RealTime timestamp
{ {
if (peakProbability[iBuf] > 0) if (peakProbability[iBuf] > 0)
{ {
double currentF0 = double currentF0 =
m_inputSampleRate * (1.0 / m_inputSampleRate * (1.0 /
YinUtil::parabolicInterpolation(yinBuffer, iBuf, yinBufferSize)); YinUtil::parabolicInterpolation(yinBuffer, iBuf, yinBufferSize));
double tempPitch = 12 * std::log(currentF0/440)/std::log(2.) + 69; double tempPitch = 12 * std::log(currentF0/440)/std::log(2.) + 69;
@ -351,10 +351,10 @@ LocalCandidatePYIN::getRemainingFeatures()
vector<float> freqSum = vector<float>(m_nCandidate); vector<float> freqSum = vector<float>(m_nCandidate);
vector<float> freqNumber = vector<float>(m_nCandidate); vector<float> freqNumber = vector<float>(m_nCandidate);
vector<float> freqMean = vector<float>(m_nCandidate); vector<float> freqMean = vector<float>(m_nCandidate);
boost::math::normal normalDist(0, 8); // semitones sd boost::math::normal normalDist(0, 8); // semitones sd
float maxNormalDist = boost::math::pdf(normalDist, 0); float maxNormalDist = boost::math::pdf(normalDist, 0);
// Viterbi-decode multiple times with different frequencies emphasised // Viterbi-decode multiple times with different frequencies emphasised
for (size_t iCandidate = 0; iCandidate < m_nCandidate; ++iCandidate) for (size_t iCandidate = 0; iCandidate < m_nCandidate; ++iCandidate)
{ {
@ -369,8 +369,8 @@ LocalCandidatePYIN::getRemainingFeatures()
float prob = 0; float prob = 0;
for (size_t iProb = 0; iProb < m_pitchProb[iFrame].size(); ++iProb) for (size_t iProb = 0; iProb < m_pitchProb[iFrame].size(); ++iProb)
{ {
pitch = m_pitchProb[iFrame][iProb].first; pitch = m_pitchProb[iFrame][iProb].first;
prob = m_pitchProb[iFrame][iProb].second * prob = m_pitchProb[iFrame][iProb].second *
boost::math::pdf(normalDist, pitch-centrePitch) / boost::math::pdf(normalDist, pitch-centrePitch) /
maxNormalDist * 2; maxNormalDist * 2;
sumProb += prob; sumProb += prob;
@ -404,13 +404,13 @@ LocalCandidatePYIN::getRemainingFeatures()
for (size_t iCandidate = 0; iCandidate < m_nCandidate; ++iCandidate) { for (size_t iCandidate = 0; iCandidate < m_nCandidate; ++iCandidate) {
for (size_t jCandidate = iCandidate+1; jCandidate < m_nCandidate; ++jCandidate) { for (size_t jCandidate = iCandidate+1; jCandidate < m_nCandidate; ++jCandidate) {
size_t countEqual = 0; size_t countEqual = 0;
for (size_t iFrame = 0; iFrame < nFrame; ++iFrame) for (size_t iFrame = 0; iFrame < nFrame; ++iFrame)
{ {
if ((pitchTracks[jCandidate][iFrame] == 0 && pitchTracks[iCandidate][iFrame] == 0) || if ((pitchTracks[jCandidate][iFrame] == 0 && pitchTracks[iCandidate][iFrame] == 0) ||
fabs(pitchTracks[iCandidate][iFrame]/pitchTracks[jCandidate][iFrame]-1)<0.01) fabs(pitchTracks[iCandidate][iFrame]/pitchTracks[jCandidate][iFrame]-1)<0.01)
countEqual++; countEqual++;
} }
// std::cerr << "proportion equal: " << (countEqual * 1.0 / nFrame) << std::endl; // std::cerr << "proportion equal: " << (countEqual * 1.0 / nFrame) << std::endl;
if (countEqual * 1.0 / nFrame > 0.8) { if (countEqual * 1.0 / nFrame > 0.8) {
if (freqNumber[iCandidate] > freqNumber[jCandidate]) { if (freqNumber[iCandidate] > freqNumber[jCandidate]) {
duplicates.push_back(jCandidate); duplicates.push_back(jCandidate);
@ -433,7 +433,7 @@ LocalCandidatePYIN::getRemainingFeatures()
{ {
bool isDuplicate = false; bool isDuplicate = false;
for (size_t i = 0; i < duplicates.size(); ++i) { for (size_t i = 0; i < duplicates.size(); ++i) {
if (duplicates[i] == iCandidate) { if (duplicates[i] == iCandidate) {
isDuplicate = true; isDuplicate = true;
break; break;
@ -446,11 +446,11 @@ LocalCandidatePYIN::getRemainingFeatures()
candidateLabels[iCandidate] = convert.str(); candidateLabels[iCandidate] = convert.str();
candidateActuals[iCandidate] = actualCandidateNumber; candidateActuals[iCandidate] = actualCandidateNumber;
// std::cerr << iCandidate << " " << actualCandidateNumber << " " << freqNumber[iCandidate] << " " << freqMean[iCandidate] << std::endl; // std::cerr << iCandidate << " " << actualCandidateNumber << " " << freqNumber[iCandidate] << " " << freqMean[iCandidate] << std::endl;
for (size_t iFrame = 0; iFrame < nFrame; ++iFrame) for (size_t iFrame = 0; iFrame < nFrame; ++iFrame)
{ {
if (pitchTracks[iCandidate][iFrame] > 0) if (pitchTracks[iCandidate][iFrame] > 0)
{ {
// featureValues[m_timestamp[iFrame]][iCandidate] = // featureValues[m_timestamp[iFrame]][iCandidate] =
// pitchTracks[iCandidate][iFrame]; // pitchTracks[iCandidate][iFrame];
outputFrequencies[iFrame].push_back(pitchTracks[iCandidate][iFrame]); outputFrequencies[iFrame].push_back(pitchTracks[iCandidate][iFrame]);
} else { } else {
@ -473,7 +473,7 @@ LocalCandidatePYIN::getRemainingFeatures()
f.values = outputFrequencies[iFrame]; f.values = outputFrequencies[iFrame];
fs[0].push_back(f); fs[0].push_back(f);
} }
// I stopped using Chris's map stuff below because I couldn't get my head around it // I stopped using Chris's map stuff below because I couldn't get my head around it
// //
// for (map<RealTime, map<int, float> >::const_iterator i = // for (map<RealTime, map<int, float> >::const_iterator i =
@ -482,7 +482,7 @@ LocalCandidatePYIN::getRemainingFeatures()
// f.hasTimestamp = true; // f.hasTimestamp = true;
// f.timestamp = i->first; // f.timestamp = i->first;
// int nextCandidate = candidateActuals.begin()->second; // int nextCandidate = candidateActuals.begin()->second;
// for (map<int, float>::const_iterator j = // for (map<int, float>::const_iterator j =
// i->second.begin(); j != i->second.end(); ++j) { // i->second.begin(); j != i->second.end(); ++j) {
// while (candidateActuals[j->first] > nextCandidate) { // while (candidateActuals[j->first] > nextCandidate) {
// f.values.push_back(0); // f.values.push_back(0);

6
libs/vamp-pyin/LocalCandidatePYIN.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -61,9 +61,9 @@ protected:
size_t m_blockSize; size_t m_blockSize;
float m_fmin; float m_fmin;
float m_fmax; float m_fmax;
mutable int m_oPitchTrackCandidates; mutable int m_oPitchTrackCandidates;
float m_threshDistr; float m_threshDistr;
float m_outputUnvoiced; float m_outputUnvoiced;
float m_preciseTime; float m_preciseTime;

2
libs/vamp-pyin/MeanFilter.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the

12
libs/vamp-pyin/MonoNote.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -38,13 +38,13 @@ MonoNote::process(const vector<vector<pair<double, double> > > pitchProb)
{ {
obsProb.push_back(hmm.calculateObsProb(pitchProb[iFrame])); obsProb.push_back(hmm.calculateObsProb(pitchProb[iFrame]));
} }
vector<double> *scale = new vector<double>(pitchProb.size()); vector<double> *scale = new vector<double>(pitchProb.size());
vector<MonoNote::FrameOutput> out; vector<MonoNote::FrameOutput> out;
vector<int> path = hmm.decodeViterbi(obsProb, scale); vector<int> path = hmm.decodeViterbi(obsProb, scale);
for (size_t iFrame = 0; iFrame < path.size(); ++iFrame) for (size_t iFrame = 0; iFrame < path.size(); ++iFrame)
{ {
double currPitch = -1; double currPitch = -1;

4
libs/vamp-pyin/MonoNote.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -28,7 +28,7 @@ class MonoNote {
public: public:
MonoNote(); MonoNote();
virtual ~MonoNote(); virtual ~MonoNote();
struct FrameOutput { struct FrameOutput {
size_t frameNumber; size_t frameNumber;
double pitch; double pitch;

35
libs/vamp-pyin/MonoNoteHMM.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -31,9 +31,9 @@ const vector<double>
MonoNoteHMM::calculateObsProb(const vector<pair<double, double> > pitchProb) MonoNoteHMM::calculateObsProb(const vector<pair<double, double> > pitchProb)
{ {
// pitchProb is a list of pairs (pitches and their probabilities) // pitchProb is a list of pairs (pitches and their probabilities)
size_t nCandidate = pitchProb.size(); size_t nCandidate = pitchProb.size();
// what is the probability of pitched // what is the probability of pitched
double pIsPitched = 0; double pIsPitched = 0;
for (size_t iCandidate = 0; iCandidate < nCandidate; ++iCandidate) for (size_t iCandidate = 0; iCandidate < nCandidate; ++iCandidate)
@ -68,8 +68,8 @@ MonoNoteHMM::calculateObsProb(const vector<pair<double, double> > pitchProb)
minDistCandidate = iCandidate; minDistCandidate = iCandidate;
} }
} }
tempProb = std::pow(minDistProb, par.yinTrust) * tempProb = std::pow(minDistProb, par.yinTrust) *
boost::math::pdf(pitchDistr[i], boost::math::pdf(pitchDistr[i],
pitchProb[minDistCandidate].first); pitchProb[minDistCandidate].first);
} else { } else {
tempProb = 1; tempProb = 1;
@ -78,12 +78,12 @@ MonoNoteHMM::calculateObsProb(const vector<pair<double, double> > pitchProb)
out[i] = tempProb; out[i] = tempProb;
} }
} }
for (size_t i = 0; i < par.n; ++i) for (size_t i = 0; i < par.n; ++i)
{ {
if (i % par.nSPP != 2) if (i % par.nSPP != 2)
{ {
if (tempProbSum > 0) if (tempProbSum > 0)
{ {
out[i] = out[i] / tempProbSum * pIsPitched; out[i] = out[i] / tempProbSum * pIsPitched;
} }
@ -106,7 +106,7 @@ MonoNoteHMM::build()
// 3-5. second-lowest pitch // 3-5. second-lowest pitch
// 3. attack state // 3. attack state
// ... // ...
// observation distributions // observation distributions
for (size_t iState = 0; iState < par.n; ++iState) for (size_t iState = 0; iState < par.n; ++iState)
{ {
@ -116,7 +116,7 @@ MonoNoteHMM::build()
// silent state starts tracking // silent state starts tracking
init.push_back(1.0/(par.nS * par.nPPS)); init.push_back(1.0/(par.nS * par.nPPS));
} else { } else {
init.push_back(0.0); init.push_back(0.0);
} }
} }
@ -128,7 +128,7 @@ MonoNoteHMM::build()
pitchDistr[index+1] = boost::math::normal(mu, par.sigmaYinPitchStable); pitchDistr[index+1] = boost::math::normal(mu, par.sigmaYinPitchStable);
pitchDistr[index+2] = boost::math::normal(mu, 1.0); // dummy pitchDistr[index+2] = boost::math::normal(mu, 1.0); // dummy
} }
boost::math::normal noteDistanceDistr(0, par.sigma2Note); boost::math::normal noteDistanceDistr(0, par.sigma2Note);
for (size_t iPitch = 0; iPitch < (par.nS * par.nPPS); ++iPitch) for (size_t iPitch = 0; iPitch < (par.nS * par.nPPS); ++iPitch)
@ -149,7 +149,7 @@ MonoNoteHMM::build()
from.push_back(index+1); from.push_back(index+1);
to.push_back(index+1); // to itself to.push_back(index+1); // to itself
transProb.push_back(par.pStableSelftrans); transProb.push_back(par.pStableSelftrans);
from.push_back(index+1); from.push_back(index+1);
to.push_back(index+2); // to silent to.push_back(index+2); // to silent
transProb.push_back(par.pStable2Silent); transProb.push_back(par.pStable2Silent);
@ -158,8 +158,7 @@ MonoNoteHMM::build()
from.push_back(index+2); from.push_back(index+2);
to.push_back(index+2); to.push_back(index+2);
transProb.push_back(par.pSilentSelftrans); transProb.push_back(par.pSilentSelftrans);
// the more complicated transitions from the silent // the more complicated transitions from the silent
double probSumSilent = 0; double probSumSilent = 0;
@ -168,17 +167,17 @@ MonoNoteHMM::build()
{ {
int fromPitch = iPitch; int fromPitch = iPitch;
int toPitch = jPitch; int toPitch = jPitch;
double semitoneDistance = double semitoneDistance =
std::abs(fromPitch - toPitch) * 1.0 / par.nPPS; std::abs(fromPitch - toPitch) * 1.0 / par.nPPS;
// if (std::fmod(semitoneDistance, 1) == 0 && semitoneDistance > par.minSemitoneDistance) // if (std::fmod(semitoneDistance, 1) == 0 && semitoneDistance > par.minSemitoneDistance)
if (semitoneDistance == 0 || if (semitoneDistance == 0 ||
(semitoneDistance > par.minSemitoneDistance (semitoneDistance > par.minSemitoneDistance
&& semitoneDistance < par.maxJump)) && semitoneDistance < par.maxJump))
{ {
size_t toIndex = jPitch * par.nSPP; // note attack index size_t toIndex = jPitch * par.nSPP; // note attack index
double tempWeightSilent = boost::math::pdf(noteDistanceDistr, double tempWeightSilent = boost::math::pdf(noteDistanceDistr,
semitoneDistance); semitoneDistance);
probSumSilent += tempWeightSilent; probSumSilent += tempWeightSilent;

2
libs/vamp-pyin/MonoNoteHMM.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the

14
libs/vamp-pyin/MonoNoteParameters.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -14,23 +14,23 @@
#include "MonoNoteParameters.h" #include "MonoNoteParameters.h"
MonoNoteParameters::MonoNoteParameters() : MonoNoteParameters::MonoNoteParameters() :
minPitch(35), minPitch(35),
nPPS(3), nPPS(3),
nS(69), nS(69),
nSPP(3), // states per pitch nSPP(3), // states per pitch
n(0), n(0),
initPi(0), initPi(0),
pAttackSelftrans(0.9), pAttackSelftrans(0.9),
pStableSelftrans(0.99), pStableSelftrans(0.99),
pStable2Silent(0.01), pStable2Silent(0.01),
pSilentSelftrans(0.9999), pSilentSelftrans(0.9999),
sigma2Note(0.7), sigma2Note(0.7),
maxJump(13), maxJump(13),
pInterSelftrans(0.0), pInterSelftrans(0.0),
priorPitchedProb(.7), priorPitchedProb(.7),
priorWeight(0.5), priorWeight(0.5),
minSemitoneDistance(.5), minSemitoneDistance(.5),
sigmaYinPitchAttack(5), sigmaYinPitchAttack(5),
sigmaYinPitchStable(0.8), sigmaYinPitchStable(0.8),
sigmaYinPitchInter(.1), sigmaYinPitchInter(.1),
yinTrust(0.1) yinTrust(0.1)

17
libs/vamp-pyin/MonoNoteParameters.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -25,17 +25,17 @@ class MonoNoteParameters
public: public:
MonoNoteParameters(); MonoNoteParameters();
virtual ~MonoNoteParameters(); virtual ~MonoNoteParameters();
// model architecture parameters // model architecture parameters
size_t minPitch; // lowest pitch in MIDI notes size_t minPitch; // lowest pitch in MIDI notes
size_t nPPS; // number of pitches per semitone size_t nPPS; // number of pitches per semitone
size_t nS; // number of semitones size_t nS; // number of semitones
size_t nSPP; // number of states per pitch size_t nSPP; // number of states per pitch
size_t n; // number of states (will be calcualted from other parameters) size_t n; // number of states (will be calcualted from other parameters)
// initial state probabilities // initial state probabilities
vector<double> initPi; vector<double> initPi;
// transition parameters // transition parameters
double pAttackSelftrans; double pAttackSelftrans;
double pStableSelftrans; double pStableSelftrans;
@ -44,18 +44,17 @@ public:
double sigma2Note; // standard deviation of next note Gaussian distribution double sigma2Note; // standard deviation of next note Gaussian distribution
double maxJump; double maxJump;
double pInterSelftrans; double pInterSelftrans;
double priorPitchedProb; double priorPitchedProb;
double priorWeight; double priorWeight;
double minSemitoneDistance; // minimum distance for a transition double minSemitoneDistance; // minimum distance for a transition
double sigmaYinPitchAttack; double sigmaYinPitchAttack;
double sigmaYinPitchStable; double sigmaYinPitchStable;
double sigmaYinPitchInter; double sigmaYinPitchInter;
double yinTrust; double yinTrust;
}; };
#endif #endif

12
libs/vamp-pyin/MonoPitch.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -40,15 +40,15 @@ MonoPitch::process(const vector<vector<pair<double, double> > > pitchProb)
{ {
obsProb.push_back(hmm.calculateObsProb(pitchProb[iFrame])); obsProb.push_back(hmm.calculateObsProb(pitchProb[iFrame]));
} }
vector<double> *scale = new vector<double>(0); vector<double> *scale = new vector<double>(0);
vector<float> out; vector<float> out;
// std::cerr << "before Viterbi decoding" << obsProb.size() << "ng" << obsProb[1].size() << std::endl; // std::cerr << "before Viterbi decoding" << obsProb.size() << "ng" << obsProb[1].size() << std::endl;
vector<int> path = hmm.decodeViterbi(obsProb, scale); vector<int> path = hmm.decodeViterbi(obsProb, scale);
// std::cerr << "after Viterbi decoding" << std::endl; // std::cerr << "after Viterbi decoding" << std::endl;
for (size_t iFrame = 0; iFrame < path.size(); ++iFrame) for (size_t iFrame = 0; iFrame < path.size(); ++iFrame)
{ {
// std::cerr << path[iFrame] << " " << hmm.m_freqs[path[iFrame]] << std::endl; // std::cerr << path[iFrame] << " " << hmm.m_freqs[path[iFrame]] << std::endl;

4
libs/vamp-pyin/MonoPitch.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -27,7 +27,7 @@ class MonoPitch {
public: public:
MonoPitch(); MonoPitch();
virtual ~MonoPitch(); virtual ~MonoPitch();
// pitchProb is a frame-wise vector carrying a vector of pitch-probability pairs // pitchProb is a frame-wise vector carrying a vector of pitch-probability pairs
const vector<float> process(const vector<vector<pair<double, double> > > pitchProb); const vector<float> process(const vector<vector<pair<double, double> > > pitchProb);
private: private:

18
libs/vamp-pyin/MonoPitchHMM.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -66,7 +66,7 @@ MonoPitchHMM::calculateObsProb(const vector<pair<double, double> > pitchProb)
oldd = d; oldd = d;
} }
} }
double probReallyPitched = m_yinTrust * probYinPitched; double probReallyPitched = m_yinTrust * probYinPitched;
// std::cerr << probReallyPitched << " " << probYinPitched << std::endl; // std::cerr << probReallyPitched << " " << probYinPitched << std::endl;
// damn, I forget what this is all about... // damn, I forget what this is all about...
@ -84,14 +84,14 @@ MonoPitchHMM::build()
{ {
// INITIAL VECTOR // INITIAL VECTOR
init = vector<double>(2*m_nPitch, 1.0 / 2*m_nPitch); init = vector<double>(2*m_nPitch, 1.0 / 2*m_nPitch);
// TRANSITIONS // TRANSITIONS
for (size_t iPitch = 0; iPitch < m_nPitch; ++iPitch) for (size_t iPitch = 0; iPitch < m_nPitch; ++iPitch)
{ {
int theoreticalMinNextPitch = static_cast<int>(iPitch)-static_cast<int>(m_transitionWidth/2); int theoreticalMinNextPitch = static_cast<int>(iPitch)-static_cast<int>(m_transitionWidth/2);
int minNextPitch = iPitch>m_transitionWidth/2 ? iPitch-m_transitionWidth/2 : 0; int minNextPitch = iPitch>m_transitionWidth/2 ? iPitch-m_transitionWidth/2 : 0;
int maxNextPitch = iPitch<m_nPitch-m_transitionWidth/2 ? iPitch+m_transitionWidth/2 : m_nPitch-1; int maxNextPitch = iPitch<m_nPitch-m_transitionWidth/2 ? iPitch+m_transitionWidth/2 : m_nPitch-1;
// WEIGHT VECTOR // WEIGHT VECTOR
double weightSum = 0; double weightSum = 0;
vector<double> weights; vector<double> weights;
@ -107,7 +107,7 @@ MonoPitchHMM::build()
} }
weightSum += weights[weights.size()-1]; weightSum += weights[weights.size()-1];
} }
// std::cerr << minNextPitch << " " << maxNextPitch << std::endl; // std::cerr << minNextPitch << " " << maxNextPitch << std::endl;
// TRANSITIONS TO CLOSE PITCH // TRANSITIONS TO CLOSE PITCH
for (size_t i = minNextPitch; i <= maxNextPitch; ++i) for (size_t i = minNextPitch; i <= maxNextPitch; ++i)
@ -124,7 +124,7 @@ MonoPitchHMM::build()
to.push_back(i+m_nPitch); to.push_back(i+m_nPitch);
transProb.push_back(weights[i-minNextPitch] / weightSum * m_selfTrans); transProb.push_back(weights[i-minNextPitch] / weightSum * m_selfTrans);
// transProb.push_back(weights[i-minNextPitch] / weightSum * 0.5); // transProb.push_back(weights[i-minNextPitch] / weightSum * 0.5);
from.push_back(iPitch+m_nPitch); from.push_back(iPitch+m_nPitch);
to.push_back(i); to.push_back(i);
transProb.push_back(weights[i-minNextPitch] / weightSum * (1-m_selfTrans)); transProb.push_back(weights[i-minNextPitch] / weightSum * (1-m_selfTrans));
@ -135,7 +135,7 @@ MonoPitchHMM::build()
// from.push_back(iPitch+m_nPitch); // from.push_back(iPitch+m_nPitch);
// to.push_back(2*m_nPitch); // to.push_back(2*m_nPitch);
// transProb.push_back(1-m_selfTrans); // transProb.push_back(1-m_selfTrans);
// TRANSITION FROM UNVOICED TO PITCH // TRANSITION FROM UNVOICED TO PITCH
// from.push_back(2*m_nPitch); // from.push_back(2*m_nPitch);
// to.push_back(iPitch+m_nPitch); // to.push_back(iPitch+m_nPitch);
@ -145,9 +145,9 @@ MonoPitchHMM::build()
// from.push_back(2*m_nPitch); // from.push_back(2*m_nPitch);
// to.push_back(2*m_nPitch); // to.push_back(2*m_nPitch);
// transProb.push_back(m_selfTrans); // transProb.push_back(m_selfTrans);
// for (size_t i = 0; i < from.size(); ++i) { // for (size_t i = 0; i < from.size(); ++i) {
// std::cerr << "P(["<< from[i] << " --> " << to[i] << "]) = " << transProb[i] << std::endl; // std::cerr << "P(["<< from[i] << " --> " << to[i] << "]) = " << transProb[i] << std::endl;
// } // }
} }

2
libs/vamp-pyin/MonoPitchHMM.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the

48
libs/vamp-pyin/PYinVamp.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -109,7 +109,7 @@ PYinVamp::getPreferredBlockSize() const
return 2048; return 2048;
} }
size_t size_t
PYinVamp::getPreferredStepSize() const PYinVamp::getPreferredStepSize() const
{ {
return 256; return 256;
@ -131,7 +131,7 @@ PYinVamp::ParameterList
PYinVamp::getParameterDescriptors() const PYinVamp::getParameterDescriptors() const
{ {
ParameterList list; ParameterList list;
ParameterDescriptor d; ParameterDescriptor d;
d.identifier = "threshdistr"; d.identifier = "threshdistr";
@ -241,7 +241,7 @@ PYinVamp::getParameter(string identifier) const
} }
void void
PYinVamp::setParameter(string identifier, float value) PYinVamp::setParameter(string identifier, float value)
{ {
if (identifier == "threshdistr") if (identifier == "threshdistr")
{ {
@ -293,7 +293,7 @@ PYinVamp::getOutputDescriptors() const
OutputList outputs; OutputList outputs;
OutputDescriptor d; OutputDescriptor d;
int outputNumber = 0; int outputNumber = 0;
d.identifier = "f0candidates"; d.identifier = "f0candidates";
@ -327,7 +327,7 @@ PYinVamp::getOutputDescriptors() const
d.hasDuration = false; d.hasDuration = false;
outputs.push_back(d); outputs.push_back(d);
m_oF0Probs = outputNumber++; m_oF0Probs = outputNumber++;
d.identifier = "voicedprob"; d.identifier = "voicedprob";
d.name = "Voiced Probability"; d.name = "Voiced Probability";
d.description = "Probability that the signal is voiced according to Probabilistic Yin."; d.description = "Probability that the signal is voiced according to Probabilistic Yin.";
@ -358,7 +358,7 @@ PYinVamp::getOutputDescriptors() const
d.hasDuration = false; d.hasDuration = false;
outputs.push_back(d); outputs.push_back(d);
m_oCandidateSalience = outputNumber++; m_oCandidateSalience = outputNumber++;
d.identifier = "smoothedpitchtrack"; d.identifier = "smoothedpitchtrack";
d.name = "Smoothed Pitch Track"; d.name = "Smoothed Pitch Track";
d.description = "."; d.description = ".";
@ -407,7 +407,7 @@ PYinVamp::initialise(size_t channels, size_t stepSize, size_t blockSize)
m_channels = channels; m_channels = channels;
m_stepSize = stepSize; m_stepSize = stepSize;
m_blockSize = blockSize; m_blockSize = blockSize;
reset(); reset();
return true; return true;
@ -415,15 +415,15 @@ PYinVamp::initialise(size_t channels, size_t stepSize, size_t blockSize)
void void
PYinVamp::reset() PYinVamp::reset()
{ {
m_yin.setThresholdDistr(m_threshDistr); m_yin.setThresholdDistr(m_threshDistr);
m_yin.setFrameSize(m_blockSize); m_yin.setFrameSize(m_blockSize);
m_yin.setFast(!m_preciseTime); m_yin.setFast(!m_preciseTime);
m_pitchProb.clear(); m_pitchProb.clear();
m_timestamp.clear(); m_timestamp.clear();
m_level.clear(); m_level.clear();
/* /*
std::cerr << "PYinVamp::reset" std::cerr << "PYinVamp::reset"
<< ", blockSize = " << m_blockSize << ", blockSize = " << m_blockSize
<< std::endl; << std::endl;
@ -437,9 +437,9 @@ PYinVamp::process(const float *const *inputBuffers, RealTime timestamp)
timestamp = timestamp + Vamp::RealTime::frame2RealTime(offset, lrintf(m_inputSampleRate)); timestamp = timestamp + Vamp::RealTime::frame2RealTime(offset, lrintf(m_inputSampleRate));
FeatureSet fs; FeatureSet fs;
float rms = 0; float rms = 0;
double *dInputBuffers = new double[m_blockSize]; double *dInputBuffers = new double[m_blockSize];
for (size_t i = 0; i < m_blockSize; ++i) { for (size_t i = 0; i < m_blockSize; ++i) {
dInputBuffers[i] = inputBuffers[0][i]; dInputBuffers[i] = inputBuffers[0][i];
@ -447,15 +447,15 @@ PYinVamp::process(const float *const *inputBuffers, RealTime timestamp)
} }
rms /= m_blockSize; rms /= m_blockSize;
rms = sqrt(rms); rms = sqrt(rms);
bool isLowAmplitude = (rms < m_lowAmp); bool isLowAmplitude = (rms < m_lowAmp);
Yin::YinOutput yo = m_yin.processProbabilisticYin(dInputBuffers); Yin::YinOutput yo = m_yin.processProbabilisticYin(dInputBuffers);
delete [] dInputBuffers; delete [] dInputBuffers;
m_level.push_back(yo.rms); m_level.push_back(yo.rms);
// First, get the things out of the way that we don't want to output // First, get the things out of the way that we don't want to output
// immediately, but instead save for later. // immediately, but instead save for later.
vector<pair<double, double> > tempPitchProb; vector<pair<double, double> > tempPitchProb;
for (size_t iCandidate = 0; iCandidate < yo.freqProb.size(); ++iCandidate) for (size_t iCandidate = 0; iCandidate < yo.freqProb.size(); ++iCandidate)
@ -483,7 +483,7 @@ PYinVamp::process(const float *const *inputBuffers, RealTime timestamp)
f.values.push_back(yo.freqProb[i].first); f.values.push_back(yo.freqProb[i].first);
} }
fs[m_oF0Candidates].push_back(f); fs[m_oF0Candidates].push_back(f);
// VOICEDPROB // VOICEDPROB
f.values.clear(); f.values.clear();
float voicedProb = 0; float voicedProb = 0;
@ -493,7 +493,7 @@ PYinVamp::process(const float *const *inputBuffers, RealTime timestamp)
voicedProb += yo.freqProb[i].second; voicedProb += yo.freqProb[i].second;
} }
fs[m_oF0Probs].push_back(f); fs[m_oF0Probs].push_back(f);
f.values.push_back(voicedProb); f.values.push_back(voicedProb);
fs[m_oVoicedProb].push_back(f); fs[m_oVoicedProb].push_back(f);
@ -517,7 +517,7 @@ PYinVamp::getRemainingFeatures()
Feature f; Feature f;
f.hasTimestamp = true; f.hasTimestamp = true;
f.hasDuration = false; f.hasDuration = false;
if (m_pitchProb.empty()) { if (m_pitchProb.empty()) {
return fs; return fs;
} }
@ -536,10 +536,10 @@ PYinVamp::getRemainingFeatures()
} else { } else {
f.values.push_back(mpOut[iFrame]); f.values.push_back(mpOut[iFrame]);
} }
fs[m_oSmoothedPitchTrack].push_back(f); fs[m_oSmoothedPitchTrack].push_back(f);
} }
// MONO-NOTE STUFF // MONO-NOTE STUFF
// std::cerr << "Mono Note Stuff" << std::endl; // std::cerr << "Mono Note Stuff" << std::endl;
MonoNote mn; MonoNote mn;
@ -555,19 +555,19 @@ PYinVamp::getRemainingFeatures()
} }
// vector<MonoNote::FrameOutput> mnOut = mn.process(m_pitchProb); // vector<MonoNote::FrameOutput> mnOut = mn.process(m_pitchProb);
vector<MonoNote::FrameOutput> mnOut = mn.process(smoothedPitch); vector<MonoNote::FrameOutput> mnOut = mn.process(smoothedPitch);
// turning feature into a note feature // turning feature into a note feature
f.hasTimestamp = true; f.hasTimestamp = true;
f.hasDuration = true; f.hasDuration = true;
f.values.clear(); f.values.clear();
int onsetFrame = 0; int onsetFrame = 0;
bool isVoiced = 0; bool isVoiced = 0;
bool oldIsVoiced = 0; bool oldIsVoiced = 0;
size_t nFrame = m_pitchProb.size(); size_t nFrame = m_pitchProb.size();
float minNoteFrames = (m_inputSampleRate*m_pruneThresh) / m_stepSize; float minNoteFrames = (m_inputSampleRate*m_pruneThresh) / m_stepSize;
std::vector<float> notePitchTrack; // collects pitches for one note at a time std::vector<float> notePitchTrack; // collects pitches for one note at a time
for (size_t iFrame = 0; iFrame < nFrame; ++iFrame) for (size_t iFrame = 0; iFrame < nFrame; ++iFrame)
{ {

4
libs/vamp-pyin/PYinVamp.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -62,7 +62,7 @@ protected:
float m_fmin; float m_fmin;
float m_fmax; float m_fmax;
Yin m_yin; Yin m_yin;
mutable int m_oF0Candidates; mutable int m_oF0Candidates;
mutable int m_oF0Probs; mutable int m_oF0Probs;
mutable int m_oVoicedProb; mutable int m_oVoicedProb;

2
libs/vamp-pyin/README Normal file
View File

@ -0,0 +1,2 @@
https://code.soundsoftware.ac.uk/projects/pyin
https://code.soundsoftware.ac.uk/attachments/download/1458/pyin-v1.1.tar.gz

26
libs/vamp-pyin/SparseHMM.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -26,9 +26,9 @@ SparseHMM::calculateObsProb(const vector<pair<double, double> > data)
return(vector<double>()); return(vector<double>());
} }
const std::vector<int> const std::vector<int>
SparseHMM::decodeViterbi(std::vector<vector<double> > obsProb, SparseHMM::decodeViterbi(std::vector<vector<double> > obsProb,
vector<double> *scale) vector<double> *scale)
{ {
if (obsProb.size() < 1) { if (obsProb.size() < 1) {
return vector<int>(); return vector<int>();
@ -36,10 +36,10 @@ SparseHMM::decodeViterbi(std::vector<vector<double> > obsProb,
size_t nState = init.size(); size_t nState = init.size();
size_t nFrame = obsProb.size(); size_t nFrame = obsProb.size();
// check for consistency // check for consistency
size_t nTrans = transProb.size(); size_t nTrans = transProb.size();
// declaring variables // declaring variables
std::vector<double> delta = std::vector<double>(nState); std::vector<double> delta = std::vector<double>(nState);
std::vector<double> oldDelta = std::vector<double>(nState); std::vector<double> oldDelta = std::vector<double>(nState);
@ -76,22 +76,22 @@ SparseHMM::decodeViterbi(std::vector<vector<double> > obsProb,
size_t toState; size_t toState;
double currentTransProb; double currentTransProb;
double currentValue; double currentValue;
// this is the "sparse" loop // this is the "sparse" loop
for (size_t iTrans = 0; iTrans < nTrans; ++iTrans) for (size_t iTrans = 0; iTrans < nTrans; ++iTrans)
{ {
fromState = from[iTrans]; fromState = from[iTrans];
toState = to[iTrans]; toState = to[iTrans];
currentTransProb = transProb[iTrans]; currentTransProb = transProb[iTrans];
currentValue = oldDelta[fromState] * currentTransProb; currentValue = oldDelta[fromState] * currentTransProb;
if (currentValue > delta[toState]) if (currentValue > delta[toState])
{ {
delta[toState] = currentValue; // will be multiplied by the right obs later! delta[toState] = currentValue; // will be multiplied by the right obs later!
psi[iFrame][toState] = fromState; psi[iFrame][toState] = fromState;
} }
} }
for (size_t jState = 0; jState < nState; ++jState) for (size_t jState = 0; jState < nState; ++jState)
{ {
delta[jState] *= obsProb[iFrame][jState]; delta[jState] *= obsProb[iFrame][jState];
@ -125,7 +125,7 @@ SparseHMM::decodeViterbi(std::vector<vector<double> > obsProb,
double currentValue = oldDelta[iState]; double currentValue = oldDelta[iState];
if (currentValue > bestValue) if (currentValue > bestValue)
{ {
bestValue = currentValue; bestValue = currentValue;
path[nFrame-1] = iState; path[nFrame-1] = iState;
} }
} }
@ -135,11 +135,11 @@ SparseHMM::decodeViterbi(std::vector<vector<double> > obsProb,
{ {
path[iFrame] = psi[iFrame+1][path[iFrame+1]]; path[iFrame] = psi[iFrame+1][path[iFrame+1]];
} }
// for (size_t iState = 0; iState < nState; ++iState) // for (size_t iState = 0; iState < nState; ++iState)
// { // {
// // std::cerr << psi[2][iState] << std::endl; // // std::cerr << psi[2][iState] << std::endl;
// } // }
return path; return path;
} }

4
libs/vamp-pyin/SparseHMM.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -24,7 +24,7 @@ class SparseHMM
{ {
public: public:
virtual const std::vector<double> calculateObsProb(const vector<pair<double, double> >); virtual const std::vector<double> calculateObsProb(const vector<pair<double, double> >);
const std::vector<int> decodeViterbi(std::vector<vector<double> > obs, const std::vector<int> decodeViterbi(std::vector<vector<double> > obs,
vector<double> *scale); vector<double> *scale);
vector<double> init; vector<double> init;
vector<size_t> from; vector<size_t> from;

24
libs/vamp-pyin/Yin.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -25,7 +25,7 @@
using std::vector; using std::vector;
Yin::Yin(size_t frameSize, size_t inputSampleRate, double thresh, bool fast) : Yin::Yin(size_t frameSize, size_t inputSampleRate, double thresh, bool fast) :
m_frameSize(frameSize), m_frameSize(frameSize),
m_inputSampleRate(inputSampleRate), m_inputSampleRate(inputSampleRate),
m_thresh(thresh), m_thresh(thresh),
@ -38,13 +38,13 @@ Yin::Yin(size_t frameSize, size_t inputSampleRate, double thresh, bool fast) :
} }
} }
Yin::~Yin() Yin::~Yin()
{ {
} }
Yin::YinOutput Yin::YinOutput
Yin::process(const double *in) const { Yin::process(const double *in) const {
double* yinBuffer = new double[m_yinBufferSize]; double* yinBuffer = new double[m_yinBufferSize];
// calculate aperiodicity function for all periods // calculate aperiodicity function for all periods
@ -55,11 +55,11 @@ Yin::process(const double *in) const {
int tau = 0; int tau = 0;
tau = YinUtil::absoluteThreshold(yinBuffer, m_yinBufferSize, m_thresh); tau = YinUtil::absoluteThreshold(yinBuffer, m_yinBufferSize, m_thresh);
double interpolatedTau; double interpolatedTau;
double aperiodicity; double aperiodicity;
double f0; double f0;
if (tau!=0) if (tau!=0)
{ {
interpolatedTau = YinUtil::parabolicInterpolation(yinBuffer, abs(tau), m_yinBufferSize); interpolatedTau = YinUtil::parabolicInterpolation(yinBuffer, abs(tau), m_yinBufferSize);
@ -78,14 +78,14 @@ Yin::process(const double *in) const {
{ {
yo.salience.push_back(yinBuffer[iBuf] < 1 ? 1-yinBuffer[iBuf] : 0); // why are the values sometimes < 0 if I don't check? yo.salience.push_back(yinBuffer[iBuf] < 1 ? 1-yinBuffer[iBuf] : 0); // why are the values sometimes < 0 if I don't check?
} }
delete [] yinBuffer; delete [] yinBuffer;
return yo; return yo;
} }
Yin::YinOutput Yin::YinOutput
Yin::processProbabilisticYin(const double *in) const { Yin::processProbabilisticYin(const double *in) const {
double* yinBuffer = new double[m_yinBufferSize]; double* yinBuffer = new double[m_yinBufferSize];
// calculate aperiodicity function for all periods // calculate aperiodicity function for all periods
@ -95,7 +95,7 @@ Yin::processProbabilisticYin(const double *in) const {
YinUtil::cumulativeDifference(yinBuffer, m_yinBufferSize); YinUtil::cumulativeDifference(yinBuffer, m_yinBufferSize);
vector<double> peakProbability = YinUtil::yinProb(yinBuffer, m_threshDistr, m_yinBufferSize); vector<double> peakProbability = YinUtil::yinProb(yinBuffer, m_threshDistr, m_yinBufferSize);
// calculate overall "probability" from peak probability // calculate overall "probability" from peak probability
double probSum = 0; double probSum = 0;
for (size_t iBin = 0; iBin < m_yinBufferSize; ++iBin) for (size_t iBin = 0; iBin < m_yinBufferSize; ++iBin)
@ -109,15 +109,15 @@ Yin::processProbabilisticYin(const double *in) const {
yo.salience.push_back(peakProbability[iBuf]); yo.salience.push_back(peakProbability[iBuf]);
if (peakProbability[iBuf] > 0) if (peakProbability[iBuf] > 0)
{ {
double currentF0 = double currentF0 =
m_inputSampleRate * (1.0 / m_inputSampleRate * (1.0 /
YinUtil::parabolicInterpolation(yinBuffer, iBuf, m_yinBufferSize)); YinUtil::parabolicInterpolation(yinBuffer, iBuf, m_yinBufferSize));
yo.freqProb.push_back(pair<double, double>(currentF0, peakProbability[iBuf])); yo.freqProb.push_back(pair<double, double>(currentF0, peakProbability[iBuf]));
} }
} }
// std::cerr << yo.freqProb.size() << std::endl; // std::cerr << yo.freqProb.size() << std::endl;
delete [] yinBuffer; delete [] yinBuffer;
return yo; return yo;
} }

11
libs/vamp-pyin/Yin.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -27,7 +27,6 @@ using std::vector;
using std::pair; using std::pair;
class Yin class Yin
{ {
public: public:
@ -40,16 +39,16 @@ public:
double rms; double rms;
vector<double> salience; vector<double> salience;
vector<pair<double, double> > freqProb; vector<pair<double, double> > freqProb;
YinOutput() : f0(0), periodicity(0), rms(0), YinOutput() : f0(0), periodicity(0), rms(0),
salience(vector<double>(0)), freqProb(vector<pair<double, double> >(0)) { } salience(vector<double>(0)), freqProb(vector<pair<double, double> >(0)) { }
YinOutput(double _f, double _p, double _r) : YinOutput(double _f, double _p, double _r) :
f0(_f), periodicity(_p), rms(_r), f0(_f), periodicity(_p), rms(_r),
salience(vector<double>(0)), freqProb(vector<pair<double, double> >(0)) { } salience(vector<double>(0)), freqProb(vector<pair<double, double> >(0)) { }
YinOutput(double _f, double _p, double _r, vector<double> _salience) : YinOutput(double _f, double _p, double _r, vector<double> _salience) :
f0(_f), periodicity(_p), rms(_r), salience(_salience), f0(_f), periodicity(_p), rms(_r), salience(_salience),
freqProb(vector<pair<double, double> >(0)) { } freqProb(vector<pair<double, double> >(0)) { }
}; };
int setThreshold(double parameter); int setThreshold(double parameter);
int setThresholdDistr(float parameter); int setThresholdDistr(float parameter);
int setFrameSize(size_t frameSize); int setFrameSize(size_t frameSize);

79
libs/vamp-pyin/YinUtil.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -21,8 +21,8 @@
#include <boost/math/distributions.hpp> #include <boost/math/distributions.hpp>
void void
YinUtil::slowDifference(const double *in, double *yinBuffer, const size_t yinBufferSize) YinUtil::slowDifference(const double *in, double *yinBuffer, const size_t yinBufferSize)
{ {
yinBuffer[0] = 0; yinBuffer[0] = 0;
double delta ; double delta ;
@ -36,19 +36,19 @@ YinUtil::slowDifference(const double *in, double *yinBuffer, const size_t yinBuf
delta = in[i+j] - in[j]; delta = in[i+j] - in[j];
yinBuffer[i] += delta * delta; yinBuffer[i] += delta * delta;
} }
} }
} }
void void
YinUtil::fastDifference(const double *in, double *yinBuffer, const size_t yinBufferSize) YinUtil::fastDifference(const double *in, double *yinBuffer, const size_t yinBufferSize)
{ {
// DECLARE AND INITIALISE // DECLARE AND INITIALISE
// initialisation of most of the arrays here was done in a separate function, // initialisation of most of the arrays here was done in a separate function,
// with all the arrays as members of the class... moved them back here. // with all the arrays as members of the class... moved them back here.
size_t frameSize = 2 * yinBufferSize; size_t frameSize = 2 * yinBufferSize;
double *audioTransformedReal = new double[frameSize]; double *audioTransformedReal = new double[frameSize];
double *audioTransformedImag = new double[frameSize]; double *audioTransformedImag = new double[frameSize];
double *nullImag = new double[frameSize]; double *nullImag = new double[frameSize];
@ -58,13 +58,13 @@ YinUtil::fastDifference(const double *in, double *yinBuffer, const size_t yinBuf
double *yinStyleACFReal = new double[frameSize]; double *yinStyleACFReal = new double[frameSize];
double *yinStyleACFImag = new double[frameSize]; double *yinStyleACFImag = new double[frameSize];
double *powerTerms = new double[yinBufferSize]; double *powerTerms = new double[yinBufferSize];
for (size_t j = 0; j < yinBufferSize; ++j) for (size_t j = 0; j < yinBufferSize; ++j)
{ {
yinBuffer[j] = 0.; // set to zero yinBuffer[j] = 0.; // set to zero
powerTerms[j] = 0.; // set to zero powerTerms[j] = 0.; // set to zero
} }
for (size_t j = 0; j < frameSize; ++j) for (size_t j = 0; j < frameSize; ++j)
{ {
nullImag[j] = 0.; nullImag[j] = 0.;
@ -76,7 +76,7 @@ YinUtil::fastDifference(const double *in, double *yinBuffer, const size_t yinBuf
yinStyleACFReal[j] = 0.; yinStyleACFReal[j] = 0.;
yinStyleACFImag[j] = 0.; yinStyleACFImag[j] = 0.;
} }
// POWER TERM CALCULATION // POWER TERM CALCULATION
// ... for the power terms in equation (7) in the Yin paper // ... for the power terms in equation (7) in the Yin paper
powerTerms[0] = 0.0; powerTerms[0] = 0.0;
@ -86,13 +86,13 @@ YinUtil::fastDifference(const double *in, double *yinBuffer, const size_t yinBuf
// now iteratively calculate all others (saves a few multiplications) // now iteratively calculate all others (saves a few multiplications)
for (size_t tau = 1; tau < yinBufferSize; ++tau) { for (size_t tau = 1; tau < yinBufferSize; ++tau) {
powerTerms[tau] = powerTerms[tau-1] - in[tau-1] * in[tau-1] + in[tau+yinBufferSize] * in[tau+yinBufferSize]; powerTerms[tau] = powerTerms[tau-1] - in[tau-1] * in[tau-1] + in[tau+yinBufferSize] * in[tau+yinBufferSize];
} }
// YIN-STYLE AUTOCORRELATION via FFT // YIN-STYLE AUTOCORRELATION via FFT
// 1. data // 1. data
Vamp::FFT::forward(frameSize, in, nullImag, audioTransformedReal, audioTransformedImag); Vamp::FFT::forward(frameSize, in, nullImag, audioTransformedReal, audioTransformedImag);
// 2. half of the data, disguised as a convolution kernel // 2. half of the data, disguised as a convolution kernel
for (size_t j = 0; j < yinBufferSize; ++j) { for (size_t j = 0; j < yinBufferSize; ++j) {
kernel[j] = in[yinBufferSize-1-j]; kernel[j] = in[yinBufferSize-1-j];
@ -105,7 +105,7 @@ YinUtil::fastDifference(const double *in, double *yinBuffer, const size_t yinBuf
yinStyleACFImag[j] = audioTransformedReal[j]*kernelTransformedImag[j] + audioTransformedImag[j]*kernelTransformedReal[j]; // imaginary yinStyleACFImag[j] = audioTransformedReal[j]*kernelTransformedImag[j] + audioTransformedImag[j]*kernelTransformedReal[j]; // imaginary
} }
Vamp::FFT::inverse(frameSize, yinStyleACFReal, yinStyleACFImag, audioTransformedReal, audioTransformedImag); Vamp::FFT::inverse(frameSize, yinStyleACFReal, yinStyleACFImag, audioTransformedReal, audioTransformedImag);
// CALCULATION OF difference function // CALCULATION OF difference function
// ... according to (7) in the Yin paper. // ... according to (7) in the Yin paper.
for (size_t j = 0; j < yinBufferSize; ++j) { for (size_t j = 0; j < yinBufferSize; ++j) {
@ -123,15 +123,15 @@ YinUtil::fastDifference(const double *in, double *yinBuffer, const size_t yinBuf
delete [] powerTerms; delete [] powerTerms;
} }
void void
YinUtil::cumulativeDifference(double *yinBuffer, const size_t yinBufferSize) YinUtil::cumulativeDifference(double *yinBuffer, const size_t yinBufferSize)
{ {
size_t tau; size_t tau;
yinBuffer[0] = 1; yinBuffer[0] = 1;
double runningSum = 0; double runningSum = 0;
for (tau = 1; tau < yinBufferSize; ++tau) { for (tau = 1; tau < yinBufferSize; ++tau) {
runningSum += yinBuffer[tau]; runningSum += yinBuffer[tau];
if (runningSum == 0) if (runningSum == 0)
@ -140,16 +140,16 @@ YinUtil::cumulativeDifference(double *yinBuffer, const size_t yinBufferSize)
} else { } else {
yinBuffer[tau] *= tau / runningSum; yinBuffer[tau] *= tau / runningSum;
} }
} }
} }
int int
YinUtil::absoluteThreshold(const double *yinBuffer, const size_t yinBufferSize, const double thresh) YinUtil::absoluteThreshold(const double *yinBuffer, const size_t yinBufferSize, const double thresh)
{ {
size_t tau; size_t tau;
size_t minTau = 0; size_t minTau = 0;
double minVal = 1000.; double minVal = 1000.;
// using Joren Six's "loop construct" from TarsosDSP // using Joren Six's "loop construct" from TarsosDSP
tau = 2; tau = 2;
while (tau < yinBufferSize) while (tau < yinBufferSize)
@ -187,7 +187,7 @@ static float single15[100] = {0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.
static float single20[100] = {0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,1.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000}; static float single20[100] = {0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,1.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000,0.00000};
std::vector<double> std::vector<double>
YinUtil::yinProb(const double *yinBuffer, const size_t prior, const size_t yinBufferSize, const size_t minTau0, const size_t maxTau0) YinUtil::yinProb(const double *yinBuffer, const size_t prior, const size_t yinBufferSize, const size_t minTau0, const size_t maxTau0)
{ {
size_t minTau = 2; size_t minTau = 2;
size_t maxTau = yinBufferSize; size_t maxTau = yinBufferSize;
@ -201,10 +201,10 @@ YinUtil::yinProb(const double *yinBuffer, const size_t prior, const size_t yinBu
std::vector<float> thresholds; std::vector<float> thresholds;
std::vector<float> distribution; std::vector<float> distribution;
std::vector<double> peakProb = std::vector<double>(yinBufferSize); std::vector<double> peakProb = std::vector<double>(yinBufferSize);
size_t nThreshold = 100; size_t nThreshold = 100;
int nThresholdInt = nThreshold; int nThresholdInt = nThreshold;
for (int i = 0; i < nThresholdInt; ++i) for (int i = 0; i < nThresholdInt; ++i)
{ {
switch (prior) { switch (prior) {
@ -237,11 +237,10 @@ YinUtil::yinProb(const double *yinBuffer, const size_t prior, const size_t yinBu
} }
thresholds.push_back(0.01 + i*0.01); thresholds.push_back(0.01 + i*0.01);
} }
int currThreshInd = nThreshold-1; int currThreshInd = nThreshold-1;
tau = minTau; tau = minTau;
// double factor = 1.0 / (0.25 * (nThresholdInt+1) * (nThresholdInt + 1)); // factor to scale down triangular weight // double factor = 1.0 / (0.25 * (nThresholdInt+1) * (nThresholdInt + 1)); // factor to scale down triangular weight
size_t minInd = 0; size_t minInd = 0;
float minVal = 42.f; float minVal = 42.f;
@ -270,7 +269,7 @@ YinUtil::yinProb(const double *yinBuffer, const size_t prior, const size_t yinBu
// { // {
// nonPeakProb -= peakProb[i]; // nonPeakProb -= peakProb[i];
// } // }
// //
// std::cerr << tau << " " << currThreshInd << " "<< thresholds[currThreshInd] << " " << distribution[currThreshInd] << std::endl; // std::cerr << tau << " " << currThreshInd << " "<< thresholds[currThreshInd] << " " << distribution[currThreshInd] << std::endl;
float sumProb = 0; float sumProb = 0;
while (tau+1 < maxTau) while (tau+1 < maxTau)
@ -300,12 +299,12 @@ YinUtil::yinProb(const double *yinBuffer, const size_t prior, const size_t yinBu
tau++; tau++;
} }
} }
if (peakProb[minInd] > 1) { if (peakProb[minInd] > 1) {
std::cerr << "WARNING: yin has prob > 1 ??? I'm returning all zeros instead." << std::endl; std::cerr << "WARNING: yin has prob > 1 ??? I'm returning all zeros instead." << std::endl;
return(std::vector<double>(yinBufferSize)); return(std::vector<double>(yinBufferSize));
} }
double nonPeakProb = 1; double nonPeakProb = 1;
if (sumProb > 0) { if (sumProb > 0) {
for (size_t i = minTau; i < maxTau; ++i) for (size_t i = minTau; i < maxTau; ++i)
@ -316,33 +315,33 @@ YinUtil::yinProb(const double *yinBuffer, const size_t prior, const size_t yinBu
} }
if (minInd > 0) if (minInd > 0)
{ {
// std::cerr << "min set " << minVal << " " << minInd << " " << nonPeakProb << std::endl; // std::cerr << "min set " << minVal << " " << minInd << " " << nonPeakProb << std::endl;
peakProb[minInd] += nonPeakProb * minWeight; peakProb[minInd] += nonPeakProb * minWeight;
} }
return peakProb; return peakProb;
} }
double double
YinUtil::parabolicInterpolation(const double *yinBuffer, const size_t tau, const size_t yinBufferSize) YinUtil::parabolicInterpolation(const double *yinBuffer, const size_t tau, const size_t yinBufferSize)
{ {
// this is taken almost literally from Joren Six's Java implementation // this is taken almost literally from Joren Six's Java implementation
if (tau == yinBufferSize) // not valid anyway. if (tau == yinBufferSize) // not valid anyway.
{ {
return static_cast<double>(tau); return static_cast<double>(tau);
} }
double betterTau = 0.0; double betterTau = 0.0;
if (tau > 0 && tau < yinBufferSize-1) { if (tau > 0 && tau < yinBufferSize-1) {
float s0, s1, s2; float s0, s1, s2;
s0 = yinBuffer[tau-1]; s0 = yinBuffer[tau-1];
s1 = yinBuffer[tau]; s1 = yinBuffer[tau];
s2 = yinBuffer[tau+1]; s2 = yinBuffer[tau+1];
double adjustment = (s2 - s0) / (2 * (2 * s1 - s2 - s0)); double adjustment = (s2 - s0) / (2 * (2 * s1 - s2 - s0));
if (abs(adjustment)>1) adjustment = 0; if (abs(adjustment)>1) adjustment = 0;
betterTau = tau + adjustment; betterTau = tau + adjustment;
} else { } else {
// std::cerr << "WARNING: can't do interpolation at the edge (tau = " << tau << "), will return un-interpolated value.\n"; // std::cerr << "WARNING: can't do interpolation at the edge (tau = " << tau << "), will return un-interpolated value.\n";
@ -351,7 +350,7 @@ YinUtil::parabolicInterpolation(const double *yinBuffer, const size_t tau, const
return betterTau; return betterTau;
} }
double double
YinUtil::sumSquare(const double *in, const size_t start, const size_t end) YinUtil::sumSquare(const double *in, const size_t start, const size_t end)
{ {
double out = 0; double out = 0;

2
libs/vamp-pyin/YinUtil.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the

28
libs/vamp-pyin/YinVamp.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -99,7 +99,7 @@ YinVamp::getPreferredBlockSize() const
return 2048; return 2048;
} }
size_t size_t
YinVamp::getPreferredStepSize() const YinVamp::getPreferredStepSize() const
{ {
return 256; return 256;
@ -121,7 +121,7 @@ YinVamp::ParameterList
YinVamp::getParameterDescriptors() const YinVamp::getParameterDescriptors() const
{ {
ParameterList list; ParameterList list;
ParameterDescriptor d; ParameterDescriptor d;
d.identifier = "yinThreshold"; d.identifier = "yinThreshold";
d.name = "Yin threshold"; d.name = "Yin threshold";
@ -132,7 +132,7 @@ YinVamp::getParameterDescriptors() const
d.defaultValue = 0.15f; d.defaultValue = 0.15f;
d.isQuantized = true; d.isQuantized = true;
d.quantizeStep = 0.025f; d.quantizeStep = 0.025f;
list.push_back(d); list.push_back(d);
d.identifier = "outputunvoiced"; d.identifier = "outputunvoiced";
@ -166,7 +166,7 @@ YinVamp::getParameter(string identifier) const
} }
void void
YinVamp::setParameter(string identifier, float value) YinVamp::setParameter(string identifier, float value)
{ {
if (identifier == "yinThreshold") if (identifier == "yinThreshold")
{ {
@ -202,7 +202,7 @@ YinVamp::getOutputDescriptors() const
OutputList outputs; OutputList outputs;
OutputDescriptor d; OutputDescriptor d;
int outputNumber = 0; int outputNumber = 0;
d.identifier = "f0"; d.identifier = "f0";
@ -285,7 +285,7 @@ YinVamp::initialise(size_t channels, size_t stepSize, size_t blockSize)
m_channels = channels; m_channels = channels;
m_stepSize = stepSize; m_stepSize = stepSize;
m_blockSize = blockSize; m_blockSize = blockSize;
reset(); reset();
return true; return true;
@ -293,10 +293,10 @@ YinVamp::initialise(size_t channels, size_t stepSize, size_t blockSize)
void void
YinVamp::reset() YinVamp::reset()
{ {
m_yin.setThreshold(m_yinParameter); m_yin.setThreshold(m_yinParameter);
m_yin.setFrameSize(m_blockSize); m_yin.setFrameSize(m_blockSize);
/* /*
std::cerr << "YinVamp::reset: yin threshold set to " << (m_yinParameter) std::cerr << "YinVamp::reset: yin threshold set to " << (m_yinParameter)
<< ", blockSize = " << m_blockSize << ", blockSize = " << m_blockSize
<< std::endl; << std::endl;
@ -308,10 +308,10 @@ YinVamp::process(const float *const *inputBuffers, RealTime timestamp)
{ {
timestamp = timestamp + Vamp::RealTime::frame2RealTime(m_blockSize/2, lrintf(m_inputSampleRate)); timestamp = timestamp + Vamp::RealTime::frame2RealTime(m_blockSize/2, lrintf(m_inputSampleRate));
FeatureSet fs; FeatureSet fs;
double *dInputBuffers = new double[m_blockSize]; double *dInputBuffers = new double[m_blockSize];
for (size_t i = 0; i < m_blockSize; ++i) dInputBuffers[i] = inputBuffers[0][i]; for (size_t i = 0; i < m_blockSize; ++i) dInputBuffers[i] = inputBuffers[0][i];
Yin::YinOutput yo = m_yin.process(dInputBuffers); Yin::YinOutput yo = m_yin.process(dInputBuffers);
// std::cerr << "f0 in YinVamp: " << yo.f0 << std::endl; // std::cerr << "f0 in YinVamp: " << yo.f0 << std::endl;
Feature f; Feature f;
@ -341,19 +341,19 @@ YinVamp::process(const float *const *inputBuffers, RealTime timestamp)
f.values.clear(); f.values.clear();
f.values.push_back(yo.rms); f.values.push_back(yo.rms);
fs[m_outNoRms].push_back(f); fs[m_outNoRms].push_back(f);
f.values.clear(); f.values.clear();
for (size_t iBin = 0; iBin < yo.salience.size(); ++iBin) for (size_t iBin = 0; iBin < yo.salience.size(); ++iBin)
{ {
f.values.push_back(yo.salience[iBin]); f.values.push_back(yo.salience[iBin]);
} }
fs[m_outNoSalience].push_back(f); fs[m_outNoSalience].push_back(f);
f.values.clear(); f.values.clear();
// f.values[0] = yo.periodicity; // f.values[0] = yo.periodicity;
f.values.push_back(yo.periodicity); f.values.push_back(yo.periodicity);
fs[m_outNoPeriodicity].push_back(f); fs[m_outNoPeriodicity].push_back(f);
delete [] dInputBuffers; delete [] dInputBuffers;
return fs; return fs;

4
libs/vamp-pyin/YinVamp.h
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the
@ -62,7 +62,7 @@ protected:
float m_fmin; float m_fmin;
float m_fmax; float m_fmax;
Yin m_yin; Yin m_yin;
mutable int m_outNoF0; mutable int m_outNoF0;
mutable int m_outNoPeriodicity; mutable int m_outNoPeriodicity;
mutable int m_outNoRms; mutable int m_outNoRms;

2
libs/vamp-pyin/libmain.cpp
View File

@ -3,7 +3,7 @@
/* /*
pYIN - A fundamental frequency estimator for monophonic audio pYIN - A fundamental frequency estimator for monophonic audio
Centre for Digital Music, Queen Mary, University of London. Centre for Digital Music, Queen Mary, University of London.
This program is free software; you can redistribute it and/or This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License as modify it under the terms of the GNU General Public License as
published by the Free Software Foundation; either version 2 of the published by the Free Software Foundation; either version 2 of the