dspark::SpectralDenoiser< T > Class Template Reference

Learn-a-profile spectral noise reduction (hiss/hum/room-tone). More...

#include <SpectralDenoiser.h>

Public Member Functions
void	prepare (const AudioSpec &spec, int fftSize=2048)
	Prepares the STFT pipeline and the per-channel bin state.
void	reset () noexcept
	Clears signal state and per-bin gain memories (keeps profile).
void	clearProfile () noexcept
	Forgets the learned noise profile (stream-owner thread).
void	setLearning (bool learning) noexcept
	While true, incoming audio trains the noise profile.
void	setReduction (T db) noexcept
	Maximum attenuation of noise bins in dB [0, 40] (default 18): the floor of the per-bin gain. Non-finite values are ignored.
void	setThreshold (T factor) noexcept
	Noise over-subtraction factor over the learned profile, in magnitude [1, 8] (default 2): the gain rule sees the learned noise scaled by it, so higher values remove more noise and more low-level signal. Non-finite values are ignored.
bool	getLearning () const noexcept
T	getReduction () const noexcept
T	getThreshold () const noexcept
int	getLatency () const noexcept
	Latency in samples (the STFT pipeline's).
std::vector< uint8_t >	getState () const
	Serializes the parameter state (the learned profile is material- dependent content, not a preset, and is intentionally not included).
bool	setState (const uint8_t *data, size_t size)
	Restores parameters from a blob (tolerant; rejects foreign ids).
void	processBlock (AudioBufferView< T > buffer) noexcept
	Processes a block in-place. Pass-through until prepare() succeeds.

Detailed Description

template<FloatType T>
class dspark::SpectralDenoiser< T >

Learn-a-profile spectral noise reduction (hiss/hum/room-tone).

Template Parameters

T	Sample type (float or double).

Definition at line 61 of file SpectralDenoiser.h.

Member Function Documentation

clearProfile()

template<FloatType T>

void dspark::SpectralDenoiser< T >::clearProfile ( )

inlinenoexcept

Forgets the learned noise profile (stream-owner thread).

Definition at line 105 of file SpectralDenoiser.h.

getLatency()

template<FloatType T>

int dspark::SpectralDenoiser< T >::getLatency ( ) const

inlinenoexcept

Latency in samples (the STFT pipeline's).

Definition at line 142 of file SpectralDenoiser.h.

getLearning()

template<FloatType T>

bool dspark::SpectralDenoiser< T >::getLearning ( ) const

inlinenoexcept

Definition at line 137 of file SpectralDenoiser.h.

getReduction()

template<FloatType T>

T dspark::SpectralDenoiser< T >::getReduction ( ) const

inlinenoexcept

Definition at line 138 of file SpectralDenoiser.h.

getState()

template<FloatType T>

std::vector< uint8_t > dspark::SpectralDenoiser< T >::getState ( ) const

inline

Serializes the parameter state (the learned profile is material- dependent content, not a preset, and is intentionally not included).

Definition at line 146 of file SpectralDenoiser.h.

getThreshold()

template<FloatType T>

T dspark::SpectralDenoiser< T >::getThreshold ( ) const

inlinenoexcept

Definition at line 139 of file SpectralDenoiser.h.

prepare()

template<FloatType T>

void dspark::SpectralDenoiser< T >::prepare ( const AudioSpec &  spec, int  fftSize = 2048  )

inline

Prepares the STFT pipeline and the per-channel bin state.

Invalid specs (non-positive/non-finite rate, block size or channel count) are ignored: the previous state is kept and an unprepared instance stays pass-through. fftSize is sanitized by the STFT engine (power of two in [4, 1 << 20]).

Parameters

spec	Audio environment specification.
fftSize	STFT size (default 2048; larger = finer hum notching).

Definition at line 77 of file SpectralDenoiser.h.

processBlock()

template<FloatType T>

void dspark::SpectralDenoiser< T >::processBlock ( AudioBufferView< T >  buffer )

inlinenoexcept

Processes a block in-place. Pass-through until prepare() succeeds.

Definition at line 169 of file SpectralDenoiser.h.

reset()

template<FloatType T>

void dspark::SpectralDenoiser< T >::reset ( )

inlinenoexcept

Clears signal state and per-bin gain memories (keeps profile).

Definition at line 95 of file SpectralDenoiser.h.

setLearning()

template<FloatType T>

void dspark::SpectralDenoiser< T >::setLearning ( bool  learning )

inlinenoexcept

While true, incoming audio trains the noise profile.

Definition at line 114 of file SpectralDenoiser.h.

setReduction()

template<FloatType T>

void dspark::SpectralDenoiser< T >::setReduction ( T  db )

inlinenoexcept

Maximum attenuation of noise bins in dB [0, 40] (default 18): the floor of the per-bin gain. Non-finite values are ignored.

Definition at line 121 of file SpectralDenoiser.h.

setState()

template<FloatType T>

bool dspark::SpectralDenoiser< T >::setState ( const uint8_t *  data, size_t  size  )

inline

Restores parameters from a blob (tolerant; rejects foreign ids).

Definition at line 157 of file SpectralDenoiser.h.

setThreshold()

template<FloatType T>

void dspark::SpectralDenoiser< T >::setThreshold ( T  factor )

inlinenoexcept

Noise over-subtraction factor over the learned profile, in magnitude [1, 8] (default 2): the gain rule sees the learned noise scaled by it, so higher values remove more noise and more low-level signal. Non-finite values are ignored.

Definition at line 131 of file SpectralDenoiser.h.

---

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

• SpectralDenoiser.h