bool DynamicLoudnessGM2002::initializeInternal(const SignalBank &input)
{
/*
* Outer-Middle ear filter
*/
//if filter coefficients have not been provided
//use spectral weighting to approximate outer and middle ear response
bool weightSpectrum = false;
if (pathToFilterCoefs_.empty())
{
weightSpectrum = true;
//should we use for useHpf for low freqs? default is true
if (middleEarFilter_ == OME::ANSIS342007_MIDDLE_EAR_HPF)
{
modules_.push_back(unique_ptr<Module> (new Butter(3, 0, 50.0)));
}
}
else
{ //otherwise, load them
//load numpy array holding the filter coefficients
cnpy::NpyArray arr = cnpy::npy_load(pathToFilterCoefs_);
Real *data = reinterpret_cast<Real*> (arr.data);
//check if filter is IIR or FIR
bool iir = false;
if(arr.shape[0]==2)
iir = true;
//load the coefficients
RealVec bCoefs, aCoefs;
for(unsigned int i=0; i<arr.shape[1];i++)
{
bCoefs.push_back(data[i]);
if(iir)
aCoefs.push_back(data[i+arr.shape[1]]);
}
//create module
if(iir)
modules_.push_back(unique_ptr<Module> (new IIR(bCoefs, aCoefs)));
else
modules_.push_back(unique_ptr<Module> (new FIR(bCoefs)));
//clean up
delete [] data;
}
/*
* Multi-resolution spectrogram
* 10 Hz to include energy caused by sidebands for frequencies near
* 20Hz but exclude DC.
* 15001 Hz so top frequencies included.
*/
RealVec bandFreqsHz {10, 80, 500, 1250, 2540, 4050, 15001};
//window spec
RealVec windowSizeSecs {0.064, 0.032, 0.016, 0.008, 0.004, 0.002};
vector<int> windowSizeSamples(6, 0);
//round to nearest sample and force to be even such that centre samples
//are aligned (using periodic Hann window)
for (int w = 0; w < 6; w++)
{
if (isSpectralResolutionDoubled_)
windowSizeSecs[w] *= 2;
windowSizeSamples[w] = (int)round(windowSizeSecs[w] * input.getFs());
windowSizeSamples[w] += windowSizeSamples[w] % 2;
}
// hop size to the nearest sample
int hopSize = round(input.getFs() / rate_);
//power spectrum
if (isHoppingGoertzelDFTUsed_)
{
compressionCriterionInCams_ = 0;
modules_.push_back(unique_ptr<Module>
(new HoppingGoertzelDFT(bandFreqsHz,
windowSizeSamples,
hopSize,
true,
true)));
//outputModules_["PowerSpectrum"] = modules_.back().get();
}
else
{
modules_.push_back(unique_ptr<Module>
(new FrameGenerator(windowSizeSamples[0],
hopSize,
isFirstSampleAtWindowCentre_)));
//windowing: Periodic hann window
modules_.push_back(unique_ptr<Module>
(new Window(Window::HANN, windowSizeSamples, true)));
modules_.push_back(unique_ptr<Module>
(new PowerSpectrum(bandFreqsHz,
windowSizeSamples,
isSpectrumSampledUniformly_)));
}
/*
* Compression
*/
if((compressionCriterionInCams_ > 0) && (isSpectrumSampledUniformly_))
{
modules_.push_back(unique_ptr<Module>
(new CompressSpectrum(compressionCriterionInCams_)));
}
/*
* Spectral weighting if necessary
*/
if (weightSpectrum)
{
if ((middleEarFilter_ != OME::NONE) || (outerEarFilter_ != OME::NONE))
{
modules_.push_back(unique_ptr<Module>
(new WeightSpectrum(middleEarFilter_, outerEarFilter_)));
}
}
int lastSpectrumIdx = modules_.size()-1;
/*
* Roex filters
*/
if(isRoexBankFast_)
{
modules_.push_back(unique_ptr<Module>
(new FastRoexBank(filterSpacingInCams_,
isExcitationPatternInterpolated_,
isInterpolationCubic_)));
}
else
{
modules_.push_back(unique_ptr<Module>
(new RoexBankANSIS342007(1.8, 38.9, filterSpacingInCams_)));
}
outputModules_["Excitation"] = modules_.back().get();
/*
* Specific loudness
*/
isBinauralInhibitionUsed_ = isBinauralInhibitionUsed_ * (input.getNEars() == 2);
modules_.push_back(unique_ptr<Module>
(new SpecificLoudnessANSIS342007(isSpecificLoudnessANSIS342007_,
isBinauralInhibitionUsed_)));
/*
* Binaural inhibition
*/
if (isBinauralInhibitionUsed_)
{
modules_.push_back(unique_ptr<Module> (new BinauralInhibitionMG2007));
}
outputModules_["SpecificLoudness"] = modules_.back().get();
/*
* Instantaneous loudness
*/
modules_.push_back(unique_ptr<Module>
(new InstantaneousLoudness(1.0, isPresentationDiotic_)));
outputModules_["InstantaneousLoudness"] = modules_.back().get();
/*
* Short-term loudness
*/
modules_.push_back(unique_ptr<Module>
(new ARAverager(attackTimeSTL_, releaseTimeSTL_)));
outputModules_["ShortTermLoudness"] = modules_.back().get();
/*
* Long-term loudness
*/
modules_.push_back(unique_ptr<Module>
(new ARAverager(attackTimeLTL_, releaseTimeLTL_)));
outputModules_["LongTermLoudness"] = modules_.back().get();
//configure targets
configureLinearTargetModuleChain();
// Masking conditions
if ((input.getNSources() > 1) && (isPartialLoudnessUsed_))
{
LOUDNESS_DEBUG(name_
<< ": Setting up modules for partial loudness...");
// Excitation transformation based on all sources
modules_.push_back(unique_ptr<Module>
(new MultiSourceRoexBank(filterSpacingInCams_)));
outputModules_["MultiSourceExcitation"] = modules_.back().get();
int moduleIdx = modules_.size() - 1;
// Push spectrum to second excitation transformation stage
Module* ptrToWeightedSpectrum = modules_[lastSpectrumIdx].get();
ptrToWeightedSpectrum -> addTargetModule
(*outputModules_["MultiSourceExcitation"]);
// Partial loudness
modules_.push_back(unique_ptr<Module>
(new SpecificPartialLoudnessMGB1997(
isSpecificLoudnessANSIS342007_,
isBinauralInhibitionUsed_)));
if (isBinauralInhibitionUsed_)
modules_.push_back(unique_ptr<Module> (new BinauralInhibitionMG2007));
outputModules_["SpecificPartialLoudness"] = modules_.back().get();
/*
* Instantaneous partial loudness
*/
modules_.push_back(unique_ptr<Module>
(new InstantaneousLoudness(1.0, isPresentationDiotic_)));
outputModules_["InstantaneousPartialLoudness"] = modules_.back().get();
/*
* Short-term partial loudness
*/
modules_.push_back(unique_ptr<Module>
(new ARAverager(attackTimeSTL_, releaseTimeSTL_)));
outputModules_["ShortTermPartialLoudness"] = modules_.back().get();
/*
* Long-term partial loudness
*/
modules_.push_back(unique_ptr<Module>
(new ARAverager(attackTimeLTL_, releaseTimeLTL_)));
outputModules_["LongTermPartialLoudness"] = modules_.back().get();
// configure targets for second (parallel) chain
configureLinearTargetModuleChain(moduleIdx);
}
return 1;
}
bool DynamicLoudnessCH2012::initializeInternal(const SignalBank &input)
{
//if filter coefficients have not been provided
//use spectral weighting to approximate outer and middle ear response
if(!pathToFilterCoefs_.empty())
{
//load numpy array holding the filter coefficients
cnpy::NpyArray arr = cnpy::npy_load(pathToFilterCoefs_);
Real *data = reinterpret_cast<Real*> (arr.data);
//check if filter is IIR or FIR
bool iir = false;
if(arr.shape[0]==2)
iir = true;
//load the coefficients
RealVec bCoefs, aCoefs;
for(unsigned int i=0; i<arr.shape[1];i++)
{
bCoefs.push_back(data[i]);
if(iir)
aCoefs.push_back(data[i+arr.shape[1]]);
}
//create module
if(iir)
{
modules_.push_back(unique_ptr<Module>
(new IIR(bCoefs, aCoefs)));
}
else
{
modules_.push_back(unique_ptr<Module>
(new FIR(bCoefs)));
}
//clean up
delete [] data;
}
/*
* Multi-resolution spectrogram
*/
RealVec bandFreqsHz {10, 80, 500, 1250, 2540, 4050, 16001};
//window spec
RealVec windowSizeSecs {0.128, 0.064, 0.032, 0.016, 0.008, 0.004};
vector<int> windowSizeSamples(6,0);
//round to nearest sample and force to be even such that centre samples
//are aligned (using periodic Hann window)
for(int w=0; w<6; w++)
{
windowSizeSamples[w] = (int)round(windowSizeSecs[w] * input.getFs());
windowSizeSamples[w] += windowSizeSamples[w] % 2;
}
// hop size to the nearest sample
int hopSize = round(input.getFs() / rate_);
//power spectrum
if (isHoppingGoertzelDFTUsed_)
{
compressionCriterionInCams_ = 0;
modules_.push_back(unique_ptr<Module>
(new HoppingGoertzelDFT(bandFreqsHz,
windowSizeSamples,
hopSize,
true,
true)));
}
else
{
modules_.push_back(unique_ptr<Module>
(new FrameGenerator(windowSizeSamples[0],
hopSize,
isFirstSampleAtWindowCentre_)));
//windowing: Periodic hann window
modules_.push_back(unique_ptr<Module>
(new Window(Window::HANN, windowSizeSamples, true)));
modules_.push_back(unique_ptr<Module>
(new PowerSpectrum(bandFreqsHz,
windowSizeSamples,
isSpectrumSampledUniformly_)));
}
/*
* Compression
*/
if((compressionCriterionInCams_ > 0) && (isSpectrumSampledUniformly_))
{
modules_.push_back(unique_ptr<Module>
(new CompressSpectrum(compressionCriterionInCams_)));
}
/*
* Spectral weighting
*/
if(pathToFilterCoefs_.empty())
{
modules_.push_back(unique_ptr<Module>
(new WeightSpectrum(OME::CHGM2011_MIDDLE_EAR, outerEarType_)));
}
/*
* Roex filters
*/
// Set up scaling factors depending on output config
Real doubleRoexBankfactor, instantaneousLoudnessFactor;
if (isSpecificLoudnessOutput_)
{
doubleRoexBankfactor = 1.53e-8;
instantaneousLoudnessFactor = 1.0;
LOUDNESS_DEBUG(name_ << ": Excitation pattern will be scaled for specific loudness");
}
else
{
doubleRoexBankfactor = 1.0;
instantaneousLoudnessFactor = 1.53e-8;
}
isBinauralInhibitionUsed_ = isBinauralInhibitionUsed_
* (input.getNEars() == 2) * isSpecificLoudnessOutput_;
if (isBinauralInhibitionUsed_)
doubleRoexBankfactor /= 0.75;
modules_.push_back(unique_ptr<Module>
(new DoubleRoexBank(1.5, 40.2,
filterSpacingInCams_,
doubleRoexBankfactor,
isExcitationPatternInterpolated_,
isInterpolationCubic_)));
/*
* Binaural inhibition
*/
if (isBinauralInhibitionUsed_)
{
modules_.push_back(unique_ptr<Module>
(new BinauralInhibitionMG2007));
}
else
{
LOUDNESS_DEBUG(name_ << ": No binaural inhibition.");
}
outputModules_["SpecificLoudness"] = modules_.back().get();
/*
* Instantaneous loudness
*/
modules_.push_back(unique_ptr<Module>
(new InstantaneousLoudness(instantaneousLoudnessFactor,
isPresentationDiotic_)));
outputModules_["InstantaneousLoudness"] = modules_.back().get();
/*
* Short-term loudness
*/
modules_.push_back(unique_ptr<Module>
(new ARAverager(attackTimeSTL_, releaseTimeSTL_)));
outputModules_["ShortTermLoudness"] = modules_.back().get();
/*
* Long-term loudness
*/
modules_.push_back(unique_ptr<Module>
(new ARAverager(attackTimeLTL_, releaseTimeLTL_)));
outputModules_["LongTermLoudness"] = modules_.back().get();
//configure targets
configureLinearTargetModuleChain();
//Option to provide PeakFollower
if (isPeakSTLFollowerUsed_)
{
modules_.push_back(unique_ptr<Module> (new PeakFollower(2.0)));
outputModules_["PeakShortTermLoudness"] = modules_.back().get();
outputModules_["ShortTermLoudness"] ->
addTargetModule (*outputModules_["PeakShortTermLoudness"]);
}
return 1;
}
