This class takes an input audio signal and a sequence of beat locations (calculated e.g. More...

#include <DownBeat.h>

Public Member Functions
	DownBeat (float originalSampleRate, size_t decimationFactor, size_t dfIncrement)
	Construct a downbeat locator that will operate on audio at the downsampled by the given decimation factor from the given original sample rate, plus beats extracted from the same audio at the given original sample rate with the given frame increment. More...

	~DownBeat ()

void	setBeatsPerBar (int bpb)

void	findDownBeats (const float *audio, size_t audioLength, const std::vector< double > &beats, std::vector< int > &downbeats)
	Estimate which beats are down-beats. More...

void	getBeatSD (std::vector< double > &beatsd) const
	Return the beat spectral difference function. More...

void	pushAudioBlock (const float *audio)
	For your downsampling convenience: call this function repeatedly with input audio blocks containing dfIncrement samples at the original sample rate, to decimate them to the downsampled rate and buffer them within the DownBeat class. More...

const float *	getBufferedAudio (size_t &length) const
	Retrieve the accumulated audio produced by pushAudioBlock calls. More...

void	resetAudioBuffer ()
	Clear any buffered downsampled audio data. More...

Private Types
typedef std::vector< int >	i_vec_t

typedef std::vector< std::vector< int > >	i_mat_t

typedef std::vector< double >	d_vec_t

typedef std::vector< std::vector< double > >	d_mat_t

Private Member Functions
void	makeDecimators ()

double	measureSpecDiff (d_vec_t oldspec, d_vec_t newspec)

Private Attributes
int	m_bpb

float	m_rate

size_t	m_factor

size_t	m_increment

Decimator *	m_decimator1

Decimator *	m_decimator2

float *	m_buffer

float *	m_decbuf

size_t	m_bufsiz

size_t	m_buffill

size_t	m_beatframesize

double *	m_beatframe

FFTReal *	m_fft

double *	m_fftRealOut

double *	m_fftImagOut

d_vec_t	m_beatsd

Detailed Description

This class takes an input audio signal and a sequence of beat locations (calculated e.g.

by TempoTrackV2) and estimates which of the beat locations are downbeats (first beat of the bar).

The input audio signal is expected to have been downsampled to a very low sampling rate (e.g. 2700Hz). A utility function for downsampling and buffering incoming block-by-block audio is provided.

Definition at line 36 of file DownBeat.h.

Member Typedef Documentation

typedef std::vector<int> DownBeat::i_vec_t

private

Definition at line 108 of file DownBeat.h.

typedef std::vector<std::vector<int> > DownBeat::i_mat_t

private

Definition at line 109 of file DownBeat.h.

typedef std::vector<double> DownBeat::d_vec_t

private

Definition at line 110 of file DownBeat.h.

typedef std::vector<std::vector<double> > DownBeat::d_mat_t

private

Definition at line 111 of file DownBeat.h.

Constructor & Destructor Documentation

DownBeat::DownBeat	(	float	originalSampleRate,
		size_t	decimationFactor,
		size_t	dfIncrement
	)

Construct a downbeat locator that will operate on audio at the downsampled by the given decimation factor from the given original sample rate, plus beats extracted from the same audio at the given original sample rate with the given frame increment.

decimationFactor must be a power of two no greater than 64, and dfIncrement must be a multiple of decimationFactor.

Definition at line 28 of file DownBeat.cpp.

References m_beatframe, m_beatframesize, m_fft, m_fftImagOut, m_fftRealOut, m_rate, and MathUtilities::nextPowerOfTwo().

DownBeat::~DownBeat ( )

Definition at line 58 of file DownBeat.cpp.

References m_beatframe, m_buffer, m_decbuf, m_decimator1, m_decimator2, m_fft, m_fftImagOut, and m_fftRealOut.

Member Function Documentation

void DownBeat::setBeatsPerBar ( int bpb )

Definition at line 71 of file DownBeat.cpp.

References m_bpb.

void DownBeat::findDownBeats	(	const float *	audio,
		size_t	audioLength,
		const std::vector< double > &	beats,
		std::vector< int > &	downbeats
	)

Estimate which beats are down-beats.

audio contains the input audio stream after downsampling, and audioLength contains the number of samples in this downsampled stream.

beats contains a series of beat positions expressed in multiples of the df increment at the audio's original sample rate, as described to the constructor.

The returned downbeat array contains a series of indices to the beats array.

Definition at line 138 of file DownBeat.cpp.

References MathUtilities::adaptiveThreshold(), FFTReal::forward(), MathUtilities::getMax(), m_beatframe, m_beatframesize, m_beatsd, m_bpb, m_factor, m_fft, m_fftImagOut, m_fftRealOut, m_increment, measureSpecDiff(), and TWO_PI.

void DownBeat::getBeatSD ( std::vector< double > & beatsd ) const

Return the beat spectral difference function.

This is calculated during findDownBeats, so this function can only be meaningfully called after that has completed. The returned vector contains one value for each of the beat times passed in to findDownBeats, less one. Each value contains the spectral difference between region prior to the beat's nominal position and the region following it.

Definition at line 293 of file DownBeat.cpp.

References m_beatsd.

void DownBeat::pushAudioBlock ( const float * audio )

For your downsampling convenience: call this function repeatedly with input audio blocks containing dfIncrement samples at the original sample rate, to decimate them to the downsampled rate and buffer them within the DownBeat class.

Call getBufferedAudio() to retrieve the results after all blocks have been processed.

Definition at line 91 of file DownBeat.cpp.

References m_buffer, m_buffill, m_bufsiz, m_decbuf, m_decimator1, m_decimator2, m_factor, m_increment, makeDecimators(), and Decimator::process().