Base class for detectors using different numerical methods. More...

#include <detectors.h>

Inheritance diagram for AbstractDetector:

Public Member Functions
	AbstractDetector (parameter_t f, parameter_t mu, parameter_t d, parameter_t sr, parameter_t detBw, parameter_t gain)

void	processAudio (discriminator_t target, const inputSample_t start, std::size_t count)
	Process audio using the numerical method appropriate to the derived class. More...

bool	searchNormalize (parameter_t searchStart, parameter_t searchEnd, const parameter_t toneDuration, const parameter_t forcing_amplitude)
	Normalise the detector frequency using an iterative scheme. More...

bool	amplitudeNormalize (const parameter_t forcing_amplitude)
	Set the detector's a and iScale attributes by profiling an ideal detector response. More...

void	scaleAmplitude ()
	Calculate amplitude scale factor.

void	generateTone (inputSample_t *tone, const std::size_t duration, const parameter_t frequency)
	Generate a sine tone. More...

parameter_t	getW (void) const
	Find the speficied characteristic frequency for this detector (post-modulation and -normalisation) More...

virtual void	reset ()=0
	Reset the internal values used to calculate z to 0. More...

Protected Member Functions
virtual void	process (discriminator_t target, const inputSample_t start, std::size_t count)=0
	This method gets overridden in derived classes to produce an unnormalised version of the required output.

template<class Archive >
void	save (Archive &archive) const
	Save the DetectorBank as a cereal archive. More...

template<class Archive >
void	load (Archive &archive)
	Load the DetectorBank from a cereal archive. More...

void	makeScaleVectors ()
	Make freq and factor vectors for given method and normalisation.

void	getScaleValue (const parameter_t fr)
	Get a scale value for a given frequency.

Static Protected Member Functions
static const parameter_t	getLyapunov (const parameter_t bw, const parameter_t amp)
	Find the first Lyapunov coefficient required for a given bandwidth, with a given forcing amplitude. More...

Protected Attributes
parameter_t	w
	Characteristic frequency.

parameter_t const	mu
	Distance from the bifurcation point.

parameter_t const	d
	Detector damping factor.

parameter_t const	sr
	Sample rate.

parameter_t const	detBw
	Detector bandwidth.

parameter_t const	gain
	Detector gain.

discriminator_t	aScale
	Amplitude scaling factor.

parameter_t	iScale
	Scaling factor for imaginary part.

parameter_t	b
	Detector's first Lyapunov coefficent.

discriminator_t	scale
	Detector's scale factor.

bool	nrml { false }
	If search normalisation has been applied.

std::vector< parameter_t >	detScaleFreqs

std::vector< discriminator_t >	detScaleFactors

Static Protected Attributes
static constexpr int	maxNormIterations { 100 }
	The maximum number of iterations to find best response during search_normalization.

static constexpr double	normConverged { 1.0028922878693671 }
	Ratio of discovered frequency to specification frequency to be considered "close enough" during seach_normalization. More...

static const std::array< std::vector< parameter_t >, 8 >	scaleFreqs

static const std::array< std::vector< discriminator_t >, 8 >	scaleFactors

Friends
class	cereal::access

template<class Archive >
void	DetectorBank::save (Archive &) const
	Permit cereal and the archive functions of the DetectorBank with which this detector is composed to read protected variables for archival purposes.

template<class Archive >
void	DetectorBank::load (Archive &)
	Permit cereal and the archive functions of the DetectorBank with which this detector is composed to write protected variables for archival purposes.

Detailed Description

Base class for detectors using different numerical methods.

The process() method must be supplied by the derived classes.

Definition at line 21 of file detectors.h.

Constructor & Destructor Documentation

◆ AbstractDetector()

AbstractDetector::AbstractDetector	(	parameter_t	f,
		parameter_t	mu,
		parameter_t	d,
		parameter_t	sr,
		parameter_t	detBw,
		parameter_t	gain
	)

Parameters

f	Detector centre frequency (Hz)
mu	Detector control parameter (should be in the region of 0)
d	Detector damping ratio
sr	Sample rate of input audio (should be 44.1kHz or 48kHz)
detBw	Detector bandwidth (Hz)
gain	Detector gain

Member Function Documentation

◆ amplitudeNormalize()

bool AbstractDetector::amplitudeNormalize ( const parameter_t forcing_amplitude )

Set the detector's a and iScale attributes by profiling an ideal detector response.

These can then be used to normalise the detector's amplitude response to the range 0-1.

Parameters

forcingAmplitude Gain that was applied to the input signal

Returns: true

Exceptions

std::string Invalid detector type while attempting amplitude normalization

◆ generateTone()

void AbstractDetector::generateTone	(	inputSample_t *	tone,
		const std::size_t	duration,
		const parameter_t	frequency
	)

Generate a sine tone.

Parameters

tone	Array for output tone
duration	Tone duration (in samples)
frequency	Frequency at which to generate tone (Hz)

◆ getLyapunov()

static const parameter_t AbstractDetector::getLyapunov	(	const parameter_t	bw,
		const parameter_t	amp
	)

staticprotected

Find the first Lyapunov coefficient required for a given bandwidth, with a given forcing amplitude.

If the desired bandwidth is given as zero, the returned value will be zero. This will create a detetor with the narrowest possible bandwidth (~1.5Hz)

Parameters

bw	Desired bandwidth
amp	Forcing amplitude

Returns: First Lyapunov coefficient

◆ getW()

parameter_t AbstractDetector::getW ( void ) const

inline

Find the speficied characteristic frequency for this detector (post-modulation and -normalisation)

Returns: Characterisic frequency in rad/s

Definition at line 104 of file detectors.h.

◆ load()

template<class Archive >

void AbstractDetector::load ( Archive & archive )

inlineprotected

Load the DetectorBank from a cereal archive.

Parameters

archive The archive from which properties are read.

Definition at line 164 of file detectors.h.

◆ processAudio()

void AbstractDetector::processAudio	(	discriminator_t *	target,
		const inputSample_t *	start,
		std::size_t	count
	)

Process audio using the numerical method appropriate to the derived class.

The derived class's process() method is called and if amplitude normalisation is required, the predetermined gain and stiffness constant is applied to the result.

Parameters

target	Pointer to output array. This should have the correct dimensions for your desired detector bank: height = number of detectors, length = length of audio input.
start	Pointer to beginning of audio
count	Number of frames to process

◆ reset()

virtual void AbstractDetector::reset ( )

pure virtual

Reset the internal values used to calculate z to 0.

This is called when DetectorBank.seek() is called. Overridden in derived classes.

Implemented in RK4Detector, and CDDetector.

◆ save()

template<class Archive >

void AbstractDetector::save ( Archive & archive ) const

inlineprotected

Save the DetectorBank as a cereal archive.

Parameters

archive The archive to which properties are written.

Definition at line 150 of file detectors.h.

◆ searchNormalize()

bool AbstractDetector::searchNormalize	(	parameter_t	searchStart,
		parameter_t	searchEnd,
		const parameter_t	toneDuration,
		const parameter_t	forcing_amplitude
	)

Normalise the detector frequency using an iterative scheme.

The lower and upper bounds of the search must be respectively lower and higher than the actual characteristic frequency of the detector. The search iterates by moving the worst performing bounding frequency half way towards the mean of the two current bounds until the maximum number of iterations or is exceeded or the specified accuracy is achieved (see maxNormIterations and normConverged).

Parameters

searchStart	Lower bound of search (ratio of specified f0)
searchEnd	Upper bound of search (ratio of specified f0)
toneDuration	Length constant test tone to be generaated

Exceptions

std::string	Invalid detector type while attempting search-normalisation
std::string	Searching for normalised charactersitc frequency: test range does not span maximum response.

Member Data Documentation

◆ normConverged

constexpr double AbstractDetector::normConverged { 1.0028922878693671 }

staticprotected

Ratio of discovered frequency to specification frequency to be considered "close enough" during seach_normalization.

1.0057929410678534 = 2^(1/120) = 10 cents; 1.0028922878693671 = 2^(1/240) = 5 cents

Definition at line 196 of file detectors.h.

The documentation for this class was generated from the following file:

/home/keziah/onsets/DetectorBank/src/detectors.h

Public Member Functions

Protected Member Functions

Static Protected Member Functions

Protected Attributes

Static Protected Attributes

Friends

Detailed Description

Constructor & Destructor Documentation

◆ AbstractDetector()

Member Function Documentation

◆ amplitudeNormalize()

◆ generateTone()

◆ getLyapunov()

◆ getW()

◆ load()

◆ processAudio()

◆ reset()

◆ save()

◆ searchNormalize()

Member Data Documentation

◆ normConverged