annotate dsp/onsets/DetectionFunction.cpp @ 298:255e431ae3d4

* Key detector: when returning key strengths, use the peak value of the
  three underlying chromagram correlations (from 36-bin chromagram)
  corresponding to each key, instead of the mean.
  Rationale: This is the same method as used when returning the key value,
  and it's nice to have the same results in both returned value and plot.
  The peak performed better than the sum with a simple test set of triads,
  so it seems reasonable to change the plot to match the key output rather
  than the other way around.

* FFT: kiss_fftr returns only the non-conjugate bins, synthesise the rest
  rather than leaving them (perhaps dangerously) undefined.  Fixes an
  uninitialised data error in chromagram that could cause garbage results
  from key detector.

* Constant Q: remove precalculated values again, I reckon they're not
  proving such a good tradeoff.

/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */

/*
    QM DSP Library

    Centre for Digital Music, Queen Mary, University of London.
    This file copyright 2005-2006 Christian Landone.
    All rights reserved.
*/

#include "DetectionFunction.h"
#include <cstring>

//////////////////////////////////////////////////////////////////////
// Construction/Destruction
//////////////////////////////////////////////////////////////////////

DetectionFunction::DetectionFunction( DFConfig Config ) :
    m_window(0)
{
    m_magHistory = NULL;
    m_phaseHistory = NULL;
    m_phaseHistoryOld = NULL;
    m_magPeaks = NULL;

    initialise( Config );
}

DetectionFunction::~DetectionFunction()
{
    deInitialise();
}


void DetectionFunction::initialise( DFConfig Config )
{
    m_dataLength = Config.frameLength;
    m_halfLength = m_dataLength/2;

    m_DFType = Config.DFType;
    m_stepSize = Config.stepSize;

    m_whiten = Config.adaptiveWhitening;
    m_whitenRelaxCoeff = Config.whiteningRelaxCoeff;
    m_whitenFloor = Config.whiteningFloor;
    if (m_whitenRelaxCoeff < 0) m_whitenRelaxCoeff = 0.9997;
    if (m_whitenFloor < 0) m_whitenFloor = 0.01;

    m_magHistory = new double[ m_halfLength ];
    memset(m_magHistory,0, m_halfLength*sizeof(double));
		
    m_phaseHistory = new double[ m_halfLength ];
    memset(m_phaseHistory,0, m_halfLength*sizeof(double));

    m_phaseHistoryOld = new double[ m_halfLength ];
    memset(m_phaseHistoryOld,0, m_halfLength*sizeof(double));

    m_magPeaks = new double[ m_halfLength ];
    memset(m_magPeaks,0, m_halfLength*sizeof(double));

    // See note in process(const double *) below
    int actualLength = MathUtilities::previousPowerOfTwo(m_dataLength);
    m_phaseVoc = new PhaseVocoder(actualLength);

    m_DFWindowedFrame = new double[ m_dataLength ];
    m_magnitude = new double[ m_halfLength ];
    m_thetaAngle = new double[ m_halfLength ];

    m_window = new Window<double>(HanningWindow, m_dataLength);
}

void DetectionFunction::deInitialise()
{
    delete [] m_magHistory ;
    delete [] m_phaseHistory ;
    delete [] m_phaseHistoryOld ;
    delete [] m_magPeaks ;

    delete m_phaseVoc;

    delete [] m_DFWindowedFrame;
    delete [] m_magnitude;
    delete [] m_thetaAngle;

    delete m_window;
}

double DetectionFunction::process( const double *TDomain )
{
    m_window->cut( TDomain, m_DFWindowedFrame );

    // Our own FFT implementation supports power-of-two sizes only.
    // If we have to use this implementation (as opposed to the
    // version of process() below that operates on frequency domain
    // data directly), we will have to use the next smallest power of
    // two from the block size.  Results may vary accordingly!

    int actualLength = MathUtilities::previousPowerOfTwo(m_dataLength);

    if (actualLength != m_dataLength) {
        // Pre-fill mag and phase vectors with zero, as the FFT output
        // will not fill the arrays
        for (int i = actualLength/2; i < m_dataLength/2; ++i) {
            m_magnitude[i] = 0;
            m_thetaAngle[0] = 0;
        }
    }

    m_phaseVoc->process(m_DFWindowedFrame, m_magnitude, m_thetaAngle);

    if (m_whiten) whiten();

    return runDF();
}

double DetectionFunction::process( const double *magnitudes, const double *phases )
{
    for (size_t i = 0; i < m_halfLength; ++i) {
        m_magnitude[i] = magnitudes[i];
        m_thetaAngle[i] = phases[i];
    }

    if (m_whiten) whiten();

    return runDF();
}

void DetectionFunction::whiten()
{
    for (unsigned int i = 0; i < m_halfLength; ++i) {
        double m = m_magnitude[i];
        if (m < m_magPeaks[i]) {
            m = m + (m_magPeaks[i] - m) * m_whitenRelaxCoeff;
        }
        if (m < m_whitenFloor) m = m_whitenFloor;
        m_magPeaks[i] = m;
        m_magnitude[i] /= m;
    }
}

double DetectionFunction::runDF()
{
    double retVal = 0;

    switch( m_DFType )
    {
    case DF_HFC:
	retVal = HFC( m_halfLength, m_magnitude);
	break;
	
    case DF_SPECDIFF:
	retVal = specDiff( m_halfLength, m_magnitude);
	break;
	
    case DF_PHASEDEV:
	retVal = phaseDev( m_halfLength, m_thetaAngle);
	break;
	
    case DF_COMPLEXSD:
	retVal = complexSD( m_halfLength, m_magnitude, m_thetaAngle);
	break;

    case DF_BROADBAND:
        retVal = broadband( m_halfLength, m_magnitude);
        break;
    }
	
    return retVal;
}

double DetectionFunction::HFC(unsigned int length, double *src)
{
    unsigned int i;
    double val = 0;

    for( i = 0; i < length; i++)
    {
	val += src[ i ] * ( i + 1);
    }
    return val;
}

double DetectionFunction::specDiff(unsigned int length, double *src)
{
    unsigned int i;
    double val = 0.0;
    double temp = 0.0;
    double diff = 0.0;

    for( i = 0; i < length; i++)
    {
	temp = fabs( (src[ i ] * src[ i ]) - (m_magHistory[ i ] * m_magHistory[ i ]) );
		
	diff= sqrt(temp);

        // (See note in phaseDev below.)

        val += diff;

	m_magHistory[ i ] = src[ i ];
    }

    return val;
}


double DetectionFunction::phaseDev(unsigned int length, double *srcPhase)
{
    unsigned int i;
    double tmpPhase = 0;
    double tmpVal = 0;
    double val = 0;

    double dev = 0;

    for( i = 0; i < length; i++)
    {
	tmpPhase = (srcPhase[ i ]- 2*m_phaseHistory[ i ]+m_phaseHistoryOld[ i ]);
	dev = MathUtilities::princarg( tmpPhase );

        // A previous version of this code only counted the value here
        // if the magnitude exceeded 0.1.  My impression is that
        // doesn't greatly improve the results for "loud" music (so
        // long as the peak picker is reasonably sophisticated), but
        // does significantly damage its ability to work with quieter
        // music, so I'm removing it and counting the result always.
        // Same goes for the spectral difference measure above.
		
        tmpVal  = fabs(dev);
        val += tmpVal ;

	m_phaseHistoryOld[ i ] = m_phaseHistory[ i ] ;
	m_phaseHistory[ i ] = srcPhase[ i ];
    }
	
	
    return val;
}


double DetectionFunction::complexSD(unsigned int length, double *srcMagnitude, double *srcPhase)
{
    unsigned int i;
    double val = 0;
    double tmpPhase = 0;
    double tmpReal = 0;
    double tmpImag = 0;
   
    double dev = 0;
    ComplexData meas = ComplexData( 0, 0 );
    ComplexData j = ComplexData( 0, 1 );

    for( i = 0; i < length; i++)
    {
	tmpPhase = (srcPhase[ i ]- 2*m_phaseHistory[ i ]+m_phaseHistoryOld[ i ]);
	dev= MathUtilities::princarg( tmpPhase );
		
	meas = m_magHistory[i] - ( srcMagnitude[ i ] * exp( j * dev) );

	tmpReal = real( meas );
	tmpImag = imag( meas );

	val += sqrt( (tmpReal * tmpReal) + (tmpImag * tmpImag) );
		
	m_phaseHistoryOld[ i ] = m_phaseHistory[ i ] ;
	m_phaseHistory[ i ] = srcPhase[ i ];
	m_magHistory[ i ] = srcMagnitude[ i ];
    }

    return val;
}

double DetectionFunction::broadband(unsigned int length, double *src)
{
    double val = 0;
    for (unsigned int i = 0; i < length; ++i) {
        double sqrmag = src[i] * src[i];
        if (m_magHistory[i] > 0.0) {
            double diff = 10.0 * log10(sqrmag / m_magHistory[i]);
            if (diff > m_dbRise) val = val + 1;
        }
        m_magHistory[i] = sqrmag;
    }
    return val;
}        

double* DetectionFunction::getSpectrumMagnitude()
{
    return m_magnitude;
}