david/librubberband
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

And per-channel formants

Browse Source
This commit is contained in:
Chris Cannam
2022-06-06 16:51:02 +01:00
parent ec7a2b1b51
commit 4fb7b0ad47
2 changed files with 52 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files

65
src/finer/R3StretcherImpl.cpp
View File

@@ -89,9 +89,6 @@ R3StretcherImpl::R3StretcherImpl(Parameters parameters,
m_resampler = std::unique_ptr<Resampler>
(new Resampler(resamplerParameters, m_parameters.channels));
m_formant = std::unique_ptr<FormantData>
(new FormantData(m_guideConfiguration.classificationFftSize));
calculateHop();
m_prevInhop = m_inhop;
@@ -383,15 +380,6 @@ R3StretcherImpl::consume()
analyseChannel(c, inhop, m_prevInhop, m_prevOuthop);
}
//!!!
/*
if (m_parameters.options & RubberBandStretcher::OptionFormantPreserved) {
m_formant->enabled = true;
analyseFormant();
} else {
m_formant->enabled = false;
}
*/
// Phase update. This is synchronised across all channels
for (auto &it : m_channelData[0]->scales) {
@@ -609,14 +597,13 @@ R3StretcherImpl::analyseChannel(int c, int inhop, int prevInhop, int prevOuthop)
}
}
//!!!
if (m_parameters.options & RubberBandStretcher::OptionFormantPreserved) {
m_formant->enabled = true;
if (c == 0) analyseFormant();
adjustFormant(c);
} else {
m_formant->enabled = false;
}
if (m_parameters.options & RubberBandStretcher::OptionFormantPreserved) {
cd->formant->enabled = true;
analyseFormant(c);
adjustFormant(c);
} else {
cd->formant->enabled = false;
}
// Use the classification scale to get a bin segmentation and
// calculate the adaptive frequency guide for this channel
@@ -652,40 +639,30 @@ R3StretcherImpl::analyseChannel(int c, int inhop, int prevInhop, int prevOuthop)
}
void
R3StretcherImpl::analyseFormant()
R3StretcherImpl::analyseFormant(int c)
{
int classify = m_guideConfiguration.classificationFftSize;
int binCount = classify/2 + 1;
int channels = m_parameters.channels;
auto &cd = m_channelData.at(c);
auto &f = *cd->formant;
auto &f = *m_formant;
v_zero(f.envelope.data(), binCount);
int fftSize = f.fftSize;
int binCount = fftSize/2 + 1;
for (int c = 0; c < channels; ++c) {
auto &cd = m_channelData.at(c);
auto &scale = cd->scales.at(classify);
for (int i = 0; i < binCount; ++i) {
f.envelope.at(i) += scale->mag.at(i) / double(channels);
}
}
auto &scale = cd->scales.at(fftSize);
auto &scaleData = m_scaleData.at(fftSize);
m_scaleData.at(classify)->fft.inverseCepstral
(f.envelope.data(), f.cepstra.data());
scaleData->fft.inverseCepstral(scale->mag.data(), f.cepstra.data());
int cutoff = int(floor(m_parameters.sampleRate / 650.0));
if (cutoff < 1) cutoff = 1;
f.cepstra[0] /= 2.0;
f.cepstra[cutoff-1] /= 2.0;
for (int i = cutoff; i < classify; ++i) {
for (int i = cutoff; i < fftSize; ++i) {
f.cepstra[i] = 0.0;
}
v_scale(f.cepstra.data(), 1.0 / double(classify), cutoff);
v_scale(f.cepstra.data(), 1.0 / double(fftSize), cutoff);
m_scaleData.at(classify)->fft.forward
(f.cepstra.data(), f.envelope.data(),
f.spare.data()); // shifted is just a spare for this one
scaleData->fft.forward(f.cepstra.data(), f.envelope.data(), f.spare.data());
v_exp(f.envelope.data(), binCount);
v_square(f.envelope.data(), binCount);
@@ -716,14 +693,14 @@ R3StretcherImpl::adjustFormant(int c)
}
*/
double targetFactor = double(m_formant->fftSize) / double(fftSize);
double targetFactor = double(cd->formant->fftSize) / double(fftSize);
double sourceFactor = targetFactor * m_pitchScale;
// double maxRatio = 60.0;
// double minRatio = 1.0 / maxRatio;
// for (int i = lowBin; i < highBin && i < formantHigh; ++i) {
for (int i = 0; i < scale->bufSize; ++i) {
double source = m_formant->envelopeAt(i * sourceFactor);
double target = m_formant->envelopeAt(i * targetFactor);
double source = cd->formant->envelopeAt(i * sourceFactor);
double target = cd->formant->envelopeAt(i * targetFactor);
if (target > 0.0) {
double ratio = source / target;
// if (ratio < minRatio) ratio = minRatio;

61
src/finer/R3StretcherImpl.h
View File

@@ -137,6 +137,33 @@ protected:
ChannelScaleData &operator=(const ChannelScaleData &) =delete;
};
struct FormantData {
bool enabled;
int fftSize;
FixedVector<double> cepstra;
FixedVector<double> envelope;
FixedVector<double> spare;
FormantData(int _fftSize) :
enabled(false),
fftSize(_fftSize),
cepstra(_fftSize, 0.0),
envelope(_fftSize/2 + 1, 0.0),
spare(_fftSize/2 + 1, 0.0) { }
double envelopeAt(double bin) const {
int b0 = int(floor(bin)), b1 = int(ceil(bin));
if (b0 < 0 || b0 > fftSize/2) {
return 0.0;
} else if (b1 == b0 || b1 > fftSize/2) {
return envelope.at(b0);
} else {
double diff = bin - double(b0);
return envelope.at(b0) * (1.0 - diff) + envelope.at(b1) * diff;
}
}
};
struct ChannelData {
std::map<int, std::shared_ptr<ChannelScaleData>> scales;
ClassificationReadaheadData readahead;
@@ -153,6 +180,7 @@ protected:
FixedVector<float> resampled;
std::unique_ptr<RingBuffer<float>> inbuf;
std::unique_ptr<RingBuffer<float>> outbuf;
std::unique_ptr<FormantData> formant;
ChannelData(BinSegmenter::Parameters segmenterParameters,
BinClassifier::Parameters classifierParameters,
int longestFftSize,
@@ -171,7 +199,8 @@ protected:
mixdown(longestFftSize, 0.f), // though it could be shorter
resampled(outRingBufferSize, 0.f),
inbuf(new RingBuffer<float>(inRingBufferSize)),
outbuf(new RingBuffer<float>(outRingBufferSize)) { }
outbuf(new RingBuffer<float>(outRingBufferSize)),
formant(new FormantData(segmenterParameters.fftSize)) { }
void reset() {
haveReadahead = false;
segmentation = BinSegmenter::Segmentation();
@@ -220,33 +249,6 @@ protected:
WindowType synthesisWindowShape(int fftSize);
int synthesisWindowLength(int fftSize);
};
struct FormantData {
bool enabled;
int fftSize;
FixedVector<double> cepstra;
FixedVector<double> envelope;
FixedVector<double> spare;
FormantData(int _fftSize) :
enabled(false),
fftSize(_fftSize),
cepstra(_fftSize, 0.0),
envelope(_fftSize/2 + 1, 0.0),
spare(_fftSize/2 + 1, 0.0) { }
double envelopeAt(double bin) const {
int b0 = int(floor(bin)), b1 = int(ceil(bin));
if (b0 < 0 || b0 > fftSize/2) {
return 0.0;
} else if (b1 == b0 || b1 > fftSize/2) {
return envelope.at(b0);
} else {
double diff = bin - double(b0);
return envelope.at(b0) * (1.0 - diff) + envelope.at(b1) * diff;
}
}
};
Parameters m_parameters;
@@ -261,7 +263,6 @@ protected:
Peak<double, std::less<double>> m_troughPicker;
std::unique_ptr<StretchCalculator> m_calculator;
std::unique_ptr<Resampler> m_resampler;
std::unique_ptr<FormantData> m_formant;
std::atomic<int> m_inhop;
int m_prevInhop;
int m_prevOuthop;
@@ -270,7 +271,7 @@ protected:
void consume();
void calculateHop();
void analyseChannel(int channel, int inhop, int prevInhop, int prevOuthop);
void analyseFormant();
void analyseFormant(int channel);
void adjustFormant(int channel);
void synthesiseChannel(int channel, int outhop);