SpectMorph::Encoder Class Reference

Encoder producing SpectMorph parametric data from sample data. More...

#include <smencoder.hh>

List of all members.

Classes
struct	Attack
Public Member Functions
	Encoder (const EncoderParams &enc_params)
void	compute_stft (GslDataHandle *dhandle, int channel, const std::vector< float > &window)
void	search_local_maxima ()
void	link_partials ()
void	validate_partials ()
void	optimize_partials (const std::vector< float > &window, int optimization_level)
void	spectral_subtract (const std::vector< float > &window)
void	approx_noise (const std::vector< float > &window)
void	compute_attack_params (const std::vector< float > &window)
void	sort_freqs ()
void	encode (GslDataHandle *dhandle, int channel, const std::vector< float > &window, int optimization_level, bool attack, bool track_sines)
void	set_time_loop (int loop_start, int loop_end)
void	save (const std::string &filename, double fundamental_freq)
Public Attributes
std::vector< AudioBlock >	audio_blocks
	current state, and end result of the encoding algorithm
Attack	optimal_attack
size_t	zero_values_at_start
size_t	sample_count
std::vector< float >	original_samples

---

Detailed Description

Encoder producing SpectMorph parametric data from sample data.

The encoder needs to perform a number of analysis steps to get from the input signal to a parametric representation (which is built in audio_blocks). At the moment, this process needs to be controlled by the caller, but a simpler interface should be added.

---

Constructor & Destructor Documentation

Encoder::Encoder

(

const EncoderParams &

enc_params

)

Constructor which initializes the Encoders parameters.

---

Member Function Documentation

void Encoder::approx_noise

(

const std::vector< float > &

window

)

This function tries to approximate the residual by a spectral envelope for a noise signal.

DEBUG CODE {

} DEBUG_CODE

void Encoder::compute_attack_params

(

const std::vector< float > &

window

)

This function computes the optimal attack parameters, by finding the optimal attack envelope (attack_start_ms and attack_end_ms) given the data.

void Encoder::compute_stft	(	GslDataHandle *	multi_channel_dhandle,
		int	channel,
		const std::vector< float > &	window
	)

This function computes the short-time-fourier-transform (STFT) of the input signal using a window to cut the individual frames out of the sample.

void Encoder::encode	(	GslDataHandle *	dhandle,
		int	channel,
		const std::vector< float > &	window,
		int	optimization_level,
		bool	attack,
		bool	track_sines
	)

This function calls all steps necessary for encoding in the right order.

Parameters:

	dhandle	a data handle containing the signal to be encoded
	window	the analysis window
	optimization_level	determines if fast (0), medium (1), or very slow (2) algorithm is used
	attack	whether to find the optimal attack parameters

void Encoder::link_partials

(

)

This function links the spectral peaks (contained in the Tracksel structure) of successive frames together by setting the prev and next pointers. It tries to minimize the frequency difference between the peaks that are linked together, while using a threshold of 5% frequency derivation.

void Encoder::optimize_partials	(	const std::vector< float > &	window,
		int	optimization_level
	)

This function reestimates the magnitudes and phases of the partials found in the previous steps.

void Encoder::save	(	const std::string &	filename,
		double	fundamental_freq
	)

This function saves the data produced by the encoder to a SpectMorph file.

void Encoder::search_local_maxima

(

)

This function searches for peaks in the frame ffts. These are stored in frame_tracksels.

void Encoder::spectral_subtract

(

const std::vector< float > &

window

)

This function subtracts the partials from the audio signal, to get the residue (remaining energy not corresponding to sine frequencies).

void Encoder::validate_partials

(

)

This function validates that the partials found by the peak linking have good quality.

---

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

• lib/smencoder.hh
• lib/smencoder.cc