A filter class for partitioned convolution. More...

Classes
struct	index_t
	Bookkeeping class. More...

Public Member Functions
	partitioned_convolution_t (unsigned int fragsize, unsigned int nchannels_in, unsigned int nchannels_out, const transfer_matrix_t &transfer)
	Create a new partitioned convolver. More...

	~partitioned_convolution_t ()
	Free fftw resource allocated in constructor. More...

mha_wave_t *	process (const mha_wave_t *s_in)
	processing More...

Public Attributes
unsigned int	fragsize
	Audio fragment size, always equal to partition size. More...

unsigned int	nchannels_in
	Number of audio input channels. More...

unsigned int	nchannels_out
	Number of audio output channels. More...

unsigned int	output_partitions
	The maximum number of partitions in any of the impulse responses. More...

unsigned int	filter_partitions
	The total number of non-zero impulse response partitions. More...

MHASignal::waveform_t	input_signal_wave
	Buffer for input signal. More...

unsigned int	current_input_signal_buffer_half_index
	A counter modulo 2. More...

MHASignal::spectrum_t	input_signal_spec
	Buffer for FFT transformed input signal. More...

MHASignal::spectrum_t	frequency_response
	Buffers for frequency response spectra of impulse response partitions. More...

std::vector< index_t >	bookkeeping
	Keeps track of input channels, output channels, impulse response partition, and delay. More...

std::vector< MHASignal::spectrum_t >	output_signal_spec
	Buffers for FFT transformed output signal. More...

unsigned int	current_output_partition_index
	A counter modulo output_partitions, indexing the "current" output partition. More...

MHASignal::waveform_t	output_signal_wave
	Buffer for the wave output signal. More...

mha_fft_t	fft
	The FFT transformer. More...

Detailed Description

A filter class for partitioned convolution.

Impulse responses are partitioned into sections of fragment size. Audio signal is convolved with every partition and delayed as needed. Convolution is done according to overlap-save. FFT length used is 2 times fragment size.

Constructor & Destructor Documentation

◆ partitioned_convolution_t()

MHAFilter::partitioned_convolution_t::partitioned_convolution_t	(	unsigned int	fragsize,
		unsigned int	nchannels_in,
		unsigned int	nchannels_out,
		const transfer_matrix_t &	transfer
	)

Create a new partitioned convolver.

Parameters

fragsize	Audio fragment size, equal to partition size.
nchannels_in	Number of input audio channels.
nchannels_out	Number of output audio channels.
transfer	A sparse matrix of impulse responses.

◆ ~partitioned_convolution_t()

MHAFilter::partitioned_convolution_t::~partitioned_convolution_t ( )

Free fftw resource allocated in constructor.

Member Function Documentation

◆ process()

mha_wave_t * MHAFilter::partitioned_convolution_t::process ( const mha_wave_t * s_in )

processing

Member Data Documentation

◆ fragsize

unsigned int MHAFilter::partitioned_convolution_t::fragsize

Audio fragment size, always equal to partition size.

◆ nchannels_in

unsigned int MHAFilter::partitioned_convolution_t::nchannels_in

Number of audio input channels.

◆ nchannels_out

unsigned int MHAFilter::partitioned_convolution_t::nchannels_out

Number of audio output channels.

◆ output_partitions

unsigned int MHAFilter::partitioned_convolution_t::output_partitions

The maximum number of partitions in any of the impulse responses.

Determines the size if the delay line.

◆ filter_partitions

unsigned int MHAFilter::partitioned_convolution_t::filter_partitions

The total number of non-zero impulse response partitions.

◆ input_signal_wave

MHASignal::waveform_t MHAFilter::partitioned_convolution_t::input_signal_wave

Buffer for input signal.

Has nchannels_in channels and fragsize*2 frames

◆ current_input_signal_buffer_half_index

unsigned int MHAFilter::partitioned_convolution_t::current_input_signal_buffer_half_index

A counter modulo 2.

Indicates the buffer half in input signal wave into which to copy the current input signal.

◆ input_signal_spec

MHASignal::spectrum_t MHAFilter::partitioned_convolution_t::input_signal_spec

Buffer for FFT transformed input signal.

Has nchannels_in channels and fragsize+1 frames (fft bins).

◆ frequency_response

MHASignal::spectrum_t MHAFilter::partitioned_convolution_t::frequency_response

Buffers for frequency response spectra of impulse response partitions.

Each "channel" contains another partition of some impulse response. The bookkeeping array is used to keep track what to do with these frequency responses. This container has filter_partitions channels and fragsize+1 frames (fft bins).

◆ bookkeeping

std::vector<index_t> MHAFilter::partitioned_convolution_t::bookkeeping

Keeps track of input channels, output channels, impulse response partition, and delay.

The index into this array is the same as the "channel" index into the frequency_response array. Array has filter_partitions entries.

◆ output_signal_spec

std::vector<MHASignal::spectrum_t> MHAFilter::partitioned_convolution_t::output_signal_spec

Buffers for FFT transformed output signal.

For each array member, Number of channels is equal to nchannels_out, number of frames (fft bins) is equal to fragsize+1. Array size is equal to output_partitions.

◆ current_output_partition_index

unsigned int MHAFilter::partitioned_convolution_t::current_output_partition_index

A counter modulo output_partitions, indexing the "current" output partition.

◆ output_signal_wave

MHASignal::waveform_t MHAFilter::partitioned_convolution_t::output_signal_wave

Buffer for the wave output signal.

Number of channels is equal to nchannels_out, number of frames is equal to fragsize

◆ fft

mha_fft_t MHAFilter::partitioned_convolution_t::fft

The FFT transformer.

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

Classes

Public Member Functions

Public Attributes