krotos::ShortTimeFourierTransform Class Reference

Apply STFT analysis to a full audio signal offline. More...

#include <ShortTimeFourierTransform.h>

Classes
struct	BinParams
struct	DFTParams

Public Types
enum	fftMode { freqMagOnly = 0 , realOnly , complex , modeNUM_MODES }
enum class	FrequencyRange { Whole , Half }

Public Member Functions
	ShortTimeFourierTransform (int winSizeSamples, int fftSizeSamples, int hopSize, WindowType winMethod, fftMode mode, int sampleRate)
std::vector< std::vector< float > >	stft (const AudioSampleBuffer &inputSignal)
Eigen::MatrixXf	processSignal (const Eigen::VectorXf inputSignal)
std::vector< float >	istft (std::vector< std::vector< float > > stftMatrix)
void	drawSpectrogram (juce::Image &image, const std::vector< std::vector< float > > &stftMatrix)
std::vector< float >	getBinsSTL (ShortTimeFourierTransform::FrequencyRange freqRange)
Eigen::VectorXf	getBins (ShortTimeFourierTransform::FrequencyRange freqRange)
	Calculates and returns the frequency bins for a given frequency range within the context of a short-time Fourier transform analysis. This function computes the frequency resolution based on the FFT size and sample rate, adjusts the frequency bins for odd or even number of points, and then selects the appropriate range of bins to return based on the specified frequency range.
std::vector< float >	getColumnsSTL ()
Eigen::VectorXf	getColumns ()
	Computes and returns the time indices for the center of each STFT window.
float	getFFTBinSize () const

Public Attributes
enum krotos::ShortTimeFourierTransform::fftMode	m_mode
struct krotos::ShortTimeFourierTransform::DFTParams	dftParams

Private Member Functions
void	NPTSinfo (float NPTS)
Eigen::VectorXf	finalGrid (const Eigen::VectorXf &w1, float Nyq, bool centerDC, ShortTimeFourierTransform::FrequencyRange freqRange)

Private Attributes
struct krotos::ShortTimeFourierTransform::BinParams	binParams
juce::dsp::FFT	m_fft
std::vector< float >	m_window
std::vector< float >	m_sigFrame
Eigen::VectorXf	sigFrame
int	m_numFrames {0}
const float	m_scalingConstant = 0.1f
int	m_inputSize {0}

Detailed Description

Apply STFT analysis to a full audio signal offline.

This class implements the STFT analysis on an audio signal. Depending on the frame size and hop size it segments the signal into overlapping windowed frames and performs an fft on each one, depending on the value fftMode has been set to. Also includes a spectrogram drawing function for debugging purposes and an istft routine. Window for now uses n+1 samples for perfect signal reconstruction(hann Window + overlap = 50%)

Member Enumeration Documentation

fftMode

enum krotos::ShortTimeFourierTransform::fftMode

enumeration choosing which FFT to use.

Enumerator
freqMagOnly
realOnly
complex
modeNUM_MODES

FrequencyRange

enum class krotos::ShortTimeFourierTransform::FrequencyRange

strong

Enumerator
Whole
Half

Constructor & Destructor Documentation

ShortTimeFourierTransform()

krotos::ShortTimeFourierTransform::ShortTimeFourierTransform	(	int	winSizeSamples,
		int	fftSizeSamples,
		int	hopSize,
		WindowType	winMethod,
		fftMode	mode,
		int	sampleRate )

Member Function Documentation

drawSpectrogram()

void krotos::ShortTimeFourierTransform::drawSpectrogram	(	juce::Image &	image,
		const std::vector< std::vector< float > > &	stftMatrix )

Draws a simple spectrogram for POC purposes. Assumes we have already produced the STFT matrix using the stft() above.

Parameters

image	the Image component whose pixels we want to colour according to the mag of the bins. TODO: write a better scaling algorithm and average the signal magnitudes.
stftMatrix	the STFT matrix in mxn dimensions where m: mth frame, n: nth bin. NOTE: you should call repaint after this so that the new Image appears.

finalGrid()

Eigen::VectorXf krotos::ShortTimeFourierTransform::finalGrid ( const Eigen::VectorXf & w1, float Nyq, bool centerDC, ShortTimeFourierTransform::FrequencyRange freqRange )

private

getBins()

Eigen::VectorXf krotos::ShortTimeFourierTransform::getBins

(

ShortTimeFourierTransform::FrequencyRange

freqRange

)

Calculates and returns the frequency bins for a given frequency range within the context of a short-time Fourier transform analysis. This function computes the frequency resolution based on the FFT size and sample rate, adjusts the frequency bins for odd or even number of points, and then selects the appropriate range of bins to return based on the specified frequency range.

Parameters

freqRange

Specifies the frequency range for which the bins should be calculated. This can be WHOLE, HALF, or any other defined range within the FrequencyRange enum.

Returns

Eigen::VectorXf A vector containing the calculated frequency bins for the specified frequency range. The size of the vector and the values within depend on the FFT size, sample rate, and the selected frequency range.

Note

The function assumes the ShortTimeFourierTransform object has been properly initialized and contains valid DFT parameters.

getBinsSTL()

std::vector< float > krotos::ShortTimeFourierTransform::getBinsSTL

(

ShortTimeFourierTransform::FrequencyRange

freqRange

)

getColumns()

Eigen::VectorXf krotos::ShortTimeFourierTransform::getColumns

(

)

Computes and returns the time indices for the center of each STFT window.

This function calculates the column offsets for the short-time Fourier transform (STFT) output based on the input size, window size, overlap, and hop size. The time indices represent the center of each STFT window, normalized by the sample rate to convert from samples to seconds. These indices can be used to understand the temporal positioning of each column in the STFT result.

Precondition

The STFT must have been computed prior to calling this function, as indicated by a non-zero input size (m_inputSize). This function relies on the assumption that m_inputSize, dftParams.winSize, dftParams.nOverlap, and dftParams.hopSize are properly initialized and represent the actual parameters used in the STFT computation.

Returns

Eigen::VectorXf A vector containing the time indices (in seconds) for the center of each STFT window. The length of the vector equals the number of columns in the STFT output.

getColumnsSTL()

std::vector< float > krotos::ShortTimeFourierTransform::getColumnsSTL

(

)

getFFTBinSize()

float krotos::ShortTimeFourierTransform::getFFTBinSize

(

)

const

istft()

std::vector< float > krotos::ShortTimeFourierTransform::istft

(

std::vector< std::vector< float > >

stftMatrix

)

Performs an inverse STFT. Assumes we have already produced the STFT matrix using the stft() above. This means that the frames retrieved from the istft are not windowed again and the mode is real only fft. TODO: If needed write version that rewindows the frames.

Parameters

stftMatrix

the STFT matrix in mxn dimensions where m: mth frame, n: nth bin. returns a vector of floats representing the reconstructed signal

NPTSinfo()

void krotos::ShortTimeFourierTransform::NPTSinfo ( float NPTS )

private

processSignal()

Eigen::MatrixXf krotos::ShortTimeFourierTransform::processSignal

(

const Eigen::VectorXf

inputSignal

)

stft()

std::vector< std::vector< float > > krotos::ShortTimeFourierTransform::stft

(

const AudioSampleBuffer &

inputSignal

)

Performs a forward STFT. Depending on the mode of the FFT we want to use it either performs a real only or a freqMagnitude only FFT. TODO: Extract information from the complex number fft.

Parameters

buffer

the whole input Sample we want to analyse frame by frame. returns a vector of vectors where each column is a different frame and each row a frequency bin. TODO:: stl removal

Member Data Documentation

binParams

struct krotos::ShortTimeFourierTransform::BinParams krotos::ShortTimeFourierTransform::binParams

private

dftParams

struct krotos::ShortTimeFourierTransform::DFTParams krotos::ShortTimeFourierTransform::dftParams

m_fft

juce::dsp::FFT krotos::ShortTimeFourierTransform::m_fft

private

m_inputSize

int krotos::ShortTimeFourierTransform::m_inputSize {0}

private

m_mode

enum krotos::ShortTimeFourierTransform::fftMode krotos::ShortTimeFourierTransform::m_mode

m_numFrames

int krotos::ShortTimeFourierTransform::m_numFrames {0}

private

m_scalingConstant

const float krotos::ShortTimeFourierTransform::m_scalingConstant = 0.1f

private

m_sigFrame

std::vector<float> krotos::ShortTimeFourierTransform::m_sigFrame

private

m_window

std::vector<float> krotos::ShortTimeFourierTransform::m_window

private

sigFrame

Eigen::VectorXf krotos::ShortTimeFourierTransform::sigFrame

private

---

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

• krotos_dsp/utilities/analysis_modules/ShortTimeFourierTransform.h
• krotos_dsp/utilities/analysis_modules/ShortTimeFourierTransform.cpp