libavcodec/aacpsy.c File Reference

AAC encoder psychoacoustic model. More...

#include "avcodec.h"
#include "aactab.h"
#include "psymodel.h"

Go to the source code of this file.

Data Structures

struct AacPsyBand

information for single band used by 3GPP TS26.403-inspired psychoacoustic model More...

struct AacPsyChannel

single/pair channel context for psychoacoustic model More...

struct AacPsyCoeffs

psychoacoustic model frame type-dependent coefficients More...

struct AacPsyContext

3GPP TS26.403-inspired psychoacoustic model specific data More...

struct PsyLamePreset

LAME psy model preset struct. More...

Defines

#define ATH_ADD   4

Functions

static float lame_calc_attack_threshold (int bitrate)

calculates the attack threshold for ABR from the above table for the LAME psy model

static void lame_window_init (AacPsyContext *ctx, AVCodecContext *avctx)

LAME psy model specific initialization.

static av_cold float calc_bark (float f)

Calculate Bark value for given line.

static av_cold float ath (float f, float add)

Calculate ATH value for given frequency.

static av_cold int psy_3gpp_init (FFPsyContext *ctx)

static float iir_filter (int in, float state[2])

IIR filter used in block switching decision.

static FFPsyWindowInfo psy_3gpp_window (FFPsyContext *ctx, const int16_t *audio, const int16_t *la, int channel, int prev_type)

Tell encoder which window types to use.

static int calc_bit_demand (AacPsyContext *ctx, float pe, int bits, int size, int short_window)

static float calc_pe_3gpp (AacPsyBand *band)

static float calc_reduction_3gpp (float a, float desired_pe, float pe, float active_lines)

static float calc_reduced_thr_3gpp (AacPsyBand *band, float min_snr, float reduction)

static void psy_3gpp_analyze (FFPsyContext *ctx, int channel, const float *coefs, const FFPsyWindowInfo *wi)

Calculate band thresholds as suggested in 3GPP TS26.403.

static av_cold void psy_3gpp_end (FFPsyContext *apc)

static void lame_apply_block_type (AacPsyChannel *ctx, FFPsyWindowInfo *wi, int uselongblock)

static FFPsyWindowInfo psy_lame_window (FFPsyContext *ctx, const int16_t *audio, const int16_t *la, int channel, int prev_type)

Variables

static const PsyLamePreset psy_abr_map []

LAME psy model preset table for ABR.

static const PsyLamePreset psy_vbr_map []

LAME psy model preset table for constant quality.

static const float psy_fir_coeffs []

LAME psy model FIR coefficient table.

static const uint8_t window_grouping [9]

window grouping information stored as bits (0 - new group, 1 - group continues)

const FFPsyModel ff_aac_psy_model

#define PSY_3GPP_THR_SPREAD_HI   1.5f

constants for 3GPP AAC psychoacoustic model

#define PSY_3GPP_THR_SPREAD_LOW   3.0f

#define PSY_3GPP_EN_SPREAD_HI_L1   2.0f

#define PSY_3GPP_EN_SPREAD_HI_L2   1.5f

#define PSY_3GPP_EN_SPREAD_HI_S   1.5f

#define PSY_3GPP_EN_SPREAD_LOW_L   3.0f

#define PSY_3GPP_EN_SPREAD_LOW_S   2.0f

#define PSY_3GPP_RPEMIN   0.01f

#define PSY_3GPP_RPELEV   2.0f

#define PSY_3GPP_C1   3.0f

#define PSY_3GPP_C2   1.3219281f

#define PSY_3GPP_C3   0.55935729f

#define PSY_SNR_1DB   7.9432821e-1f

#define PSY_SNR_25DB   3.1622776e-3f

#define PSY_3GPP_SAVE_SLOPE_L   -0.46666667f

#define PSY_3GPP_SAVE_SLOPE_S   -0.36363637f

#define PSY_3GPP_SAVE_ADD_L   -0.84285712f

#define PSY_3GPP_SAVE_ADD_S   -0.75f

#define PSY_3GPP_SPEND_SLOPE_L   0.66666669f

#define PSY_3GPP_SPEND_SLOPE_S   0.81818181f

#define PSY_3GPP_SPEND_ADD_L   -0.35f

#define PSY_3GPP_SPEND_ADD_S   -0.26111111f

#define PSY_3GPP_CLIP_LO_L   0.2f

#define PSY_3GPP_CLIP_LO_S   0.2f

#define PSY_3GPP_CLIP_HI_L   0.95f

#define PSY_3GPP_CLIP_HI_S   0.75f

#define PSY_3GPP_AH_THR_LONG   0.5f

#define PSY_3GPP_AH_THR_SHORT   0.63f

#define PSY_3GPP_BITS_TO_PE(bits)   ((bits) * 1.18f)

#define PSY_LAME_FIR_LEN   21

LAME psy model FIR order.

#define AAC_BLOCK_SIZE_LONG   1024

long block size

#define AAC_BLOCK_SIZE_SHORT   128

short block size

#define AAC_NUM_BLOCKS_SHORT   8

number of blocks in a short sequence

#define PSY_LAME_NUM_SUBBLOCKS   3

Number of sub-blocks in each short block.

enum { PSY_3GPP_AH_NONE, PSY_3GPP_AH_INACTIVE, PSY_3GPP_AH_ACTIVE }

Detailed Description

AAC encoder psychoacoustic model.

Definition in file aacpsy.c.

Define Documentation

#define AAC_BLOCK_SIZE_LONG 1024

long block size

Definition at line 90 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define AAC_BLOCK_SIZE_SHORT 128

short block size

Definition at line 91 of file aacpsy.c.

#define AAC_NUM_BLOCKS_SHORT 8

number of blocks in a short sequence

Definition at line 92 of file aacpsy.c.

Referenced by lame_window_init(), and psy_lame_window().

#define ATH_ADD 4

Definition at line 275 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_3GPP_AH_THR_LONG 0.5f

Definition at line 77 of file aacpsy.c.

#define PSY_3GPP_AH_THR_SHORT 0.63f

Definition at line 78 of file aacpsy.c.

#define PSY_3GPP_BITS_TO_PE ( bits ) ((bits) * 1.18f)

Definition at line 86 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_3GPP_C1 3.0f

Definition at line 57 of file aacpsy.c.

Referenced by calc_pe_3gpp().

#define PSY_3GPP_C2 1.3219281f

Definition at line 58 of file aacpsy.c.

Referenced by calc_pe_3gpp().

#define PSY_3GPP_C3 0.55935729f

Definition at line 59 of file aacpsy.c.

Referenced by calc_pe_3gpp().

#define PSY_3GPP_CLIP_HI_L 0.95f

Definition at line 74 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_CLIP_HI_S 0.75f

Definition at line 75 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_CLIP_LO_L 0.2f

Definition at line 72 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_CLIP_LO_S 0.2f

Definition at line 73 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_EN_SPREAD_HI_L1 2.0f

Definition at line 44 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_3GPP_EN_SPREAD_HI_L2 1.5f

Definition at line 46 of file aacpsy.c.

#define PSY_3GPP_EN_SPREAD_HI_S 1.5f

Definition at line 48 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_3GPP_EN_SPREAD_LOW_L 3.0f

Definition at line 50 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_3GPP_EN_SPREAD_LOW_S 2.0f

Definition at line 52 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_3GPP_RPELEV 2.0f

Definition at line 55 of file aacpsy.c.

#define PSY_3GPP_RPEMIN 0.01f

Definition at line 54 of file aacpsy.c.

#define PSY_3GPP_SAVE_ADD_L -0.84285712f

Definition at line 66 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_SAVE_ADD_S -0.75f

Definition at line 67 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_SAVE_SLOPE_L -0.46666667f

Definition at line 64 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_SAVE_SLOPE_S -0.36363637f

Definition at line 65 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_SPEND_ADD_L -0.35f

Definition at line 70 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_SPEND_ADD_S -0.26111111f

Definition at line 71 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_SPEND_SLOPE_L 0.66666669f

Definition at line 68 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_SPEND_SLOPE_S 0.81818181f

Definition at line 69 of file aacpsy.c.

Referenced by calc_bit_demand().

#define PSY_3GPP_THR_SPREAD_HI 1.5f

constants for 3GPP AAC psychoacoustic model

Definition at line 41 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_3GPP_THR_SPREAD_LOW 3.0f

Definition at line 42 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_LAME_FIR_LEN 21

LAME psy model FIR order.

Definition at line 89 of file aacpsy.c.

#define PSY_LAME_NUM_SUBBLOCKS 3

Number of sub-blocks in each short block.

Definition at line 93 of file aacpsy.c.

Referenced by lame_window_init().

#define PSY_SNR_1DB 7.9432821e-1f

Definition at line 61 of file aacpsy.c.

Referenced by psy_3gpp_init().

#define PSY_SNR_25DB 3.1622776e-3f

Definition at line 62 of file aacpsy.c.

Referenced by psy_3gpp_init().

Enumeration Type Documentation

anonymous enum

Enumerator:

PSY_3GPP_AH_NONE
PSY_3GPP_AH_INACTIVE
PSY_3GPP_AH_ACTIVE

Definition at line 80 of file aacpsy.c.

Function Documentation

static av_cold float ath	(	float	f,
		float	add
	)			`[static]`

Calculate ATH value for given frequency.

Borrowed from Lame.

Definition at line 280 of file aacpsy.c.

Referenced by psy_3gpp_init().

static av_cold float calc_bark ( float f ) [static]

Calculate Bark value for given line.

Definition at line 270 of file aacpsy.c.

Referenced by psy_3gpp_init().

static int calc_bit_demand	(	AacPsyContext *	ctx,
		float	pe,
		int	bits,
		int	size,
		int	short_window
	)			`[static]`

Definition at line 466 of file aacpsy.c.

static float calc_pe_3gpp ( AacPsyBand * band ) [static]

Definition at line 499 of file aacpsy.c.

static float calc_reduced_thr_3gpp	(	AacPsyBand *	band,
		float	min_snr,
		float	reduction
	)			`[static]`

Definition at line 533 of file aacpsy.c.

static float calc_reduction_3gpp	(	float	a,
		float	desired_pe,
		float	pe,
		float	active_lines
	)			`[static]`

Definition at line 522 of file aacpsy.c.

static float iir_filter	(	int	in,
		float	state[2]
	)			`[static]`

IIR filter used in block switching decision.

Definition at line 359 of file aacpsy.c.

Referenced by psy_3gpp_window().

static void lame_apply_block_type	(	AacPsyChannel *	ctx,
		FFPsyWindowInfo *	wi,
		int	uselongblock
	)			`[static]`

Definition at line 750 of file aacpsy.c.

static float lame_calc_attack_threshold ( int bitrate ) [static]

calculates the attack threshold for ABR from the above table for the LAME psy model

Definition at line 221 of file aacpsy.c.

Referenced by lame_window_init().

static void lame_window_init	(	AacPsyContext *	ctx,
		AVCodecContext *	avctx
	)			`[static]`

LAME psy model specific initialization.

Definition at line 251 of file aacpsy.c.

Referenced by psy_3gpp_init().

static void psy_3gpp_analyze	(	FFPsyContext *	ctx,
		int	channel,
		const float *	coefs,
		const FFPsyWindowInfo *	wi
	)			`[static]`

Calculate band thresholds as suggested in 3GPP TS26.403.

Definition at line 559 of file aacpsy.c.

static av_cold void psy_3gpp_end ( FFPsyContext * apc ) [static]

Definition at line 743 of file aacpsy.c.

static av_cold int psy_3gpp_init ( FFPsyContext * ctx ) [static]

Definition at line 289 of file aacpsy.c.

static FFPsyWindowInfo psy_3gpp_window	(	FFPsyContext *	ctx,
		const int16_t *	audio,
		const int16_t *	la,
		int	channel,
		int	prev_type
	)			`[static]`

Tell encoder which window types to use.

See also:: 3GPP TS26.403 5.4.1 "Blockswitching"

Definition at line 380 of file aacpsy.c.

static FFPsyWindowInfo psy_lame_window	(	FFPsyContext *	ctx,
		const int16_t *	audio,
		const int16_t *	la,
		int	channel,
		int	prev_type
	)			`[static]`

Definition at line 768 of file aacpsy.c.

Variable Documentation

const FFPsyModel ff_aac_psy_model

Initial value:

{
    .name    = "3GPP TS 26.403-inspired model",
    .init    = psy_3gpp_init,
    .window  = psy_lame_window,
    .analyze = psy_3gpp_analyze,
    .end     = psy_3gpp_end,
}

Definition at line 917 of file aacpsy.c.

const PsyLamePreset psy_abr_map[] [static]

Initial value:

 {


    {  8,  6.60},
    { 16,  6.60},
    { 24,  6.60},
    { 32,  6.60},
    { 40,  6.60},
    { 48,  6.60},
    { 56,  6.60},
    { 64,  6.40},
    { 80,  6.00},
    { 96,  5.60},
    {112,  5.20},
    {128,  5.20},
    {160,  5.20}
}

LAME psy model preset table for ABR.

Definition at line 173 of file aacpsy.c.

const float psy_fir_coeffs[] [static]

Initial value:

 {
    -8.65163e-18 * 2, -0.00851586 * 2, -6.74764e-18 * 2, 0.0209036 * 2,
    -3.36639e-17 * 2, -0.0438162 * 2,  -1.54175e-17 * 2, 0.0931738 * 2,
    -5.52212e-17 * 2, -0.313819 * 2
}

LAME psy model FIR coefficient table.

Definition at line 212 of file aacpsy.c.

const PsyLamePreset psy_vbr_map[] [static]

Initial value:

 {

    { 0,  4.20},
    { 1,  4.20},
    { 2,  4.20},
    { 3,  4.20},
    { 4,  4.20},
    { 5,  4.20},
    { 6,  4.20},
    { 7,  4.20},
    { 8,  4.20},
    { 9,  4.20},
    {10,  4.20}
}

LAME psy model preset table for constant quality.

Definition at line 194 of file aacpsy.c.

const uint8_t window_grouping[9] [static]

Initial value:

 {
    0xB6, 0x6C, 0xD8, 0xB2, 0x66, 0xC6, 0x96, 0x36, 0x36
}

window grouping information stored as bits (0 - new group, 1 - group continues)

Definition at line 372 of file aacpsy.c.

Referenced by psy_3gpp_window().


Data Structures
struct	AacPsyBand
	information for single band used by 3GPP TS26.403-inspired psychoacoustic model More...
struct	AacPsyChannel
	single/pair channel context for psychoacoustic model More...
struct	AacPsyCoeffs
	psychoacoustic model frame type-dependent coefficients More...
struct	AacPsyContext
	3GPP TS26.403-inspired psychoacoustic model specific data More...
struct	PsyLamePreset
	LAME psy model preset struct. More...
Defines
#define	ATH_ADD 4
Functions
static float	lame_calc_attack_threshold (int bitrate)
	calculates the attack threshold for ABR from the above table for the LAME psy model
static void	lame_window_init (AacPsyContext ctx, AVCodecContext avctx)
	LAME psy model specific initialization.
static av_cold float	calc_bark (float f)
	Calculate Bark value for given line.
static av_cold float	ath (float f, float add)
	Calculate ATH value for given frequency.
static av_cold int	psy_3gpp_init (FFPsyContext *ctx)
static float	iir_filter (int in, float state[2])
	IIR filter used in block switching decision.
static FFPsyWindowInfo	psy_3gpp_window (FFPsyContext ctx, const int16_t audio, const int16_t *la, int channel, int prev_type)
	Tell encoder which window types to use.
static int	calc_bit_demand (AacPsyContext *ctx, float pe, int bits, int size, int short_window)
static float	calc_pe_3gpp (AacPsyBand *band)
static float	calc_reduction_3gpp (float a, float desired_pe, float pe, float active_lines)
static float	calc_reduced_thr_3gpp (AacPsyBand *band, float min_snr, float reduction)
static void	psy_3gpp_analyze (FFPsyContext ctx, int channel, const float coefs, const FFPsyWindowInfo *wi)
	Calculate band thresholds as suggested in 3GPP TS26.403.
static av_cold void	psy_3gpp_end (FFPsyContext *apc)
static void	lame_apply_block_type (AacPsyChannel ctx, FFPsyWindowInfo wi, int uselongblock)
static FFPsyWindowInfo	psy_lame_window (FFPsyContext ctx, const int16_t audio, const int16_t *la, int channel, int prev_type)
Variables
static const PsyLamePreset	psy_abr_map []
	LAME psy model preset table for ABR.
static const PsyLamePreset	psy_vbr_map []
	LAME psy model preset table for constant quality.
static const float	psy_fir_coeffs []
	LAME psy model FIR coefficient table.
static const uint8_t	window_grouping [9]
	window grouping information stored as bits (0 - new group, 1 - group continues)
const FFPsyModel	ff_aac_psy_model
#define	PSY_3GPP_THR_SPREAD_HI 1.5f
	constants for 3GPP AAC psychoacoustic model
#define	PSY_3GPP_THR_SPREAD_LOW 3.0f
#define	PSY_3GPP_EN_SPREAD_HI_L1 2.0f
#define	PSY_3GPP_EN_SPREAD_HI_L2 1.5f
#define	PSY_3GPP_EN_SPREAD_HI_S 1.5f
#define	PSY_3GPP_EN_SPREAD_LOW_L 3.0f
#define	PSY_3GPP_EN_SPREAD_LOW_S 2.0f
#define	PSY_3GPP_RPEMIN 0.01f
#define	PSY_3GPP_RPELEV 2.0f
#define	PSY_3GPP_C1 3.0f
#define	PSY_3GPP_C2 1.3219281f
#define	PSY_3GPP_C3 0.55935729f
#define	PSY_SNR_1DB 7.9432821e-1f
#define	PSY_SNR_25DB 3.1622776e-3f
#define	PSY_3GPP_SAVE_SLOPE_L -0.46666667f
#define	PSY_3GPP_SAVE_SLOPE_S -0.36363637f
#define	PSY_3GPP_SAVE_ADD_L -0.84285712f
#define	PSY_3GPP_SAVE_ADD_S -0.75f
#define	PSY_3GPP_SPEND_SLOPE_L 0.66666669f
#define	PSY_3GPP_SPEND_SLOPE_S 0.81818181f
#define	PSY_3GPP_SPEND_ADD_L -0.35f
#define	PSY_3GPP_SPEND_ADD_S -0.26111111f
#define	PSY_3GPP_CLIP_LO_L 0.2f
#define	PSY_3GPP_CLIP_LO_S 0.2f
#define	PSY_3GPP_CLIP_HI_L 0.95f
#define	PSY_3GPP_CLIP_HI_S 0.75f
#define	PSY_3GPP_AH_THR_LONG 0.5f
#define	PSY_3GPP_AH_THR_SHORT 0.63f
#define	PSY_3GPP_BITS_TO_PE(bits) ((bits) * 1.18f)
#define	PSY_LAME_FIR_LEN 21
	LAME psy model FIR order.
#define	AAC_BLOCK_SIZE_LONG 1024
	long block size
#define	AAC_BLOCK_SIZE_SHORT 128
	short block size
#define	AAC_NUM_BLOCKS_SHORT 8
	number of blocks in a short sequence
#define	PSY_LAME_NUM_SUBBLOCKS 3
	Number of sub-blocks in each short block.
enum	{ PSY_3GPP_AH_NONE, PSY_3GPP_AH_INACTIVE, PSY_3GPP_AH_ACTIVE }