bool
MF0UA::initialise(size_t channels, size_t stepSize, size_t blockSize)
{
if (channels < getMinChannelCount() ||
channels > getMaxChannelCount()) return false;
// Real initialisation work goes here!
m_stepSize = stepSize;
m_blockSize = blockSize;
// Initialize spectruminfo
initializeSpectrumInfo();
// Memory allocation for window
window = (double*) malloc(sizeof(double)*(spectruminfo.winsize));
// Creation of the Hanning window
Hanning(window,spectruminfo.N);
// Bands generation (Only when the algorithm is onset-based)
if (algorithm==2)
{
generatebands(spectruminfo.first_band_freq, spectruminfo.samplerate/2, spectralbands, spectruminfo.freq_resolution);
spectruminfo.numbands=spectralbands.size();
// Do not compute differences for the first frame
resolutiondiff=(getPreferredBlockSize()/2)/getPreferredStepSize();
}
// This is the first frame
firstframe=true;
n_time=0;
return true;
}
SpeechMusicSegmenter::OutputList
SpeechMusicSegmenter::getOutputDescriptors() const
{
OutputList list;
OutputDescriptor segmentation;
segmentation.identifier = "segmentation";
segmentation.name = "Segmentation";
segmentation.description = "Segmentation";
segmentation.unit = "segment-type";
segmentation.hasFixedBinCount = true;
segmentation.binCount = 1;
segmentation.hasKnownExtents = true;
segmentation.minValue = 0;
segmentation.maxValue = 2;
segmentation.isQuantized = true;
segmentation.quantizeStep = 1;
segmentation.sampleType = OutputDescriptor::VariableSampleRate;
segmentation.sampleRate = m_inputSampleRate / getPreferredStepSize();
OutputDescriptor skewness;
skewness.identifier = "skewness";
skewness.name = "Detection function";
skewness.description = "Detection function";
skewness.unit = "segment-type";
skewness.hasFixedBinCount = true;
skewness.binCount = 1;
skewness.hasKnownExtents = true;
skewness.minValue = 0;
skewness.maxValue = 2;
skewness.isQuantized = true;
skewness.quantizeStep = 1;
skewness.sampleType = OutputDescriptor::VariableSampleRate;
skewness.sampleRate = m_inputSampleRate / getPreferredStepSize();
list.push_back(segmentation);
list.push_back(skewness);
return list;
}
TonalChangeDetect::OutputList TonalChangeDetect::getOutputDescriptors() const
{
OutputList list;
OutputDescriptor hc;
hc.identifier = "tcstransform";
hc.name = "Transform to 6D Tonal Content Space";
hc.unit = "";
hc.description = "Representation of content in a six-dimensional tonal space";
hc.hasFixedBinCount = true;
hc.binCount = 6;
hc.hasKnownExtents = true;
hc.minValue = -1.0;
hc.maxValue = 1.0;
hc.isQuantized = false;
hc.sampleType = OutputDescriptor::OneSamplePerStep;
OutputDescriptor d;
d.identifier = "tcfunction";
d.name = "Tonal Change Detection Function";
d.unit = "";
d.description = "Estimate of the likelihood of a tonal change occurring within each spectral frame";
d.minValue = 0;
d.minValue = 2;
d.hasFixedBinCount = true;
d.binCount = 1;
d.hasKnownExtents = false;
d.isQuantized = false;
d.sampleType = OutputDescriptor::VariableSampleRate;
double dStepSecs = double(getPreferredStepSize()) / m_inputSampleRate;
d.sampleRate = 1.0f / dStepSecs;
OutputDescriptor changes;
changes.identifier = "changepositions";
changes.name = "Tonal Change Positions";
changes.unit = "";
changes.description = "Estimated locations of tonal changes";
changes.hasFixedBinCount = true;
changes.binCount = 0;
changes.hasKnownExtents = false;
changes.isQuantized = false;
changes.sampleType = OutputDescriptor::VariableSampleRate;
changes.sampleRate = 1.0 / dStepSecs;
list.push_back(hc);
list.push_back(d);
list.push_back(changes);
return list;
}
void
MF0UA::initializeSpectrumInfo()
{
spectruminfo.first_band_freq=kMINBANDFREQ; // Minimum frequency for band processing
spectruminfo.samplerate= sr;
spectruminfo.N=getPreferredBlockSize();
spectruminfo.res=kRES; // 4 by default, for zero padding
spectruminfo.percentage=100-(getPreferredStepSize()/(float)(spectruminfo.N))*100;
spectruminfo.freq_resolution=(double)(spectruminfo.samplerate)/((double)(spectruminfo.N)*spectruminfo.res);
spectruminfo.winsize= spectruminfo.N*spectruminfo.res;
spectruminfo.firstsample=(int)round(spectruminfo.W/spectruminfo.freq_resolution); // Convert bandwidth (Hz->Samples)
if (spectruminfo.percentage!=0) spectruminfo.time_resolution= (1.0-((double)spectruminfo.percentage/100.0))* (spectruminfo.N) / spectruminfo.samplerate;
else spectruminfo.time_resolution=(double)(spectruminfo.N)/(double)(spectruminfo.samplerate);
spectruminfo.min_sample=(int)floor(spectruminfo.first_band_freq/spectruminfo.freq_resolution);
}
