FFmpeg  4.0
Data Structures | Macros | Functions | Variables
aacpsy.c File Reference

AAC encoder psychoacoustic model. More...

#include "libavutil/attributes.h"
#include "libavutil/ffmath.h"
#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...
 

Macros

#define ATH_ADD   4
 

Functions

static float lame_calc_attack_threshold (int bitrate)
 Calculate the ABR attack threshold from the above LAME psymodel table. More...
 
static av_cold void lame_window_init (AacPsyContext *ctx, AVCodecContext *avctx)
 LAME psy model specific initialization. More...
 
static av_cold float calc_bark (float f)
 Calculate Bark value for given line. More...
 
static av_cold float ath (float f, float add)
 Calculate ATH value for given frequency. More...
 
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. More...
 
static av_unused 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. More...
 
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 calc_thr_3gpp (const FFPsyWindowInfo *wi, const int num_bands, AacPsyChannel *pch, const uint8_t *band_sizes, const float *coefs, const int cutoff)
 
static void psy_hp_filter (const float *firbuf, float *hpfsmpl, const float *psy_fir_coeffs)
 
static void psy_3gpp_analyze_channel (FFPsyContext *ctx, int channel, const float *coefs, const FFPsyWindowInfo *wi)
 Calculate band thresholds as suggested in 3GPP TS26.403. More...
 
static void psy_3gpp_analyze (FFPsyContext *ctx, int channel, const float **coeffs, const FFPsyWindowInfo *wi)
 
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 float *audio, const float *la, int channel, int prev_type)
 

Variables

static const PsyLamePreset psy_abr_map []
 LAME psy model preset table for ABR. More...
 
static const PsyLamePreset psy_vbr_map []
 LAME psy model preset table for constant quality. More...
 
static const float psy_fir_coeffs []
 LAME psy model FIR coefficient table. More...
 
static const uint8_t window_grouping [9]
 window grouping information stored as bits (0 - new group, 1 - group continues) More...
 
const FFPsyModel ff_aac_psy_model
 
#define PSY_3GPP_THR_SPREAD_HI   1.5f
 constants for 3GPP AAC psychoacoustic model More...
 
#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 /* log2(8) */
 
#define PSY_3GPP_C2   1.3219281f /* log2(2.5) */
 
#define PSY_3GPP_C3   0.55935729f /* 1 - C2 / C1 */
 
#define PSY_SNR_1DB   7.9432821e-1f /* -1dB */
 
#define PSY_SNR_25DB   3.1622776e-3f /* -25dB */
 
#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_PE_FORGET_SLOPE   511
 
#define PSY_3GPP_BITS_TO_PE(bits)   ((bits) * 1.18f)
 
#define PSY_3GPP_PE_TO_BITS(bits)   ((bits) / 1.18f)
 
#define PSY_LAME_FIR_LEN   21
 LAME psy model FIR order. More...
 
#define AAC_BLOCK_SIZE_LONG   1024
 long block size More...
 
#define AAC_BLOCK_SIZE_SHORT   128
 short block size More...
 
#define AAC_NUM_BLOCKS_SHORT   8
 number of blocks in a short sequence More...
 
#define PSY_LAME_NUM_SUBBLOCKS   3
 Number of sub-blocks in each short block. More...
 
enum  { PSY_3GPP_AH_NONE, PSY_3GPP_AH_INACTIVE, PSY_3GPP_AH_ACTIVE }
 

Detailed Description

AAC encoder psychoacoustic model.

Definition in file aacpsy.c.

Macro Definition Documentation

◆ PSY_3GPP_THR_SPREAD_HI

#define PSY_3GPP_THR_SPREAD_HI   1.5f

constants for 3GPP AAC psychoacoustic model

Definition at line 44 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_THR_SPREAD_LOW

#define PSY_3GPP_THR_SPREAD_LOW   3.0f

Definition at line 45 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_EN_SPREAD_HI_L1

#define PSY_3GPP_EN_SPREAD_HI_L1   2.0f

Definition at line 47 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_EN_SPREAD_HI_L2

#define PSY_3GPP_EN_SPREAD_HI_L2   1.5f

Definition at line 49 of file aacpsy.c.

◆ PSY_3GPP_EN_SPREAD_HI_S

#define PSY_3GPP_EN_SPREAD_HI_S   1.5f

Definition at line 51 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_EN_SPREAD_LOW_L

#define PSY_3GPP_EN_SPREAD_LOW_L   3.0f

Definition at line 53 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_EN_SPREAD_LOW_S

#define PSY_3GPP_EN_SPREAD_LOW_S   2.0f

Definition at line 55 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_RPEMIN

#define PSY_3GPP_RPEMIN   0.01f

Definition at line 57 of file aacpsy.c.

◆ PSY_3GPP_RPELEV

#define PSY_3GPP_RPELEV   2.0f

Definition at line 58 of file aacpsy.c.

◆ PSY_3GPP_C1

#define PSY_3GPP_C1   3.0f /* log2(8) */

Definition at line 60 of file aacpsy.c.

Referenced by calc_pe_3gpp().

◆ PSY_3GPP_C2

#define PSY_3GPP_C2   1.3219281f /* log2(2.5) */

Definition at line 61 of file aacpsy.c.

Referenced by calc_pe_3gpp().

◆ PSY_3GPP_C3

#define PSY_3GPP_C3   0.55935729f /* 1 - C2 / C1 */

Definition at line 62 of file aacpsy.c.

Referenced by calc_pe_3gpp().

◆ PSY_SNR_1DB

#define PSY_SNR_1DB   7.9432821e-1f /* -1dB */

Definition at line 64 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_SNR_25DB

#define PSY_SNR_25DB   3.1622776e-3f /* -25dB */

Definition at line 65 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_SAVE_SLOPE_L

#define PSY_3GPP_SAVE_SLOPE_L   -0.46666667f

Definition at line 67 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_SAVE_SLOPE_S

#define PSY_3GPP_SAVE_SLOPE_S   -0.36363637f

Definition at line 68 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_SAVE_ADD_L

#define PSY_3GPP_SAVE_ADD_L   -0.84285712f

Definition at line 69 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_SAVE_ADD_S

#define PSY_3GPP_SAVE_ADD_S   -0.75f

Definition at line 70 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_SPEND_SLOPE_L

#define PSY_3GPP_SPEND_SLOPE_L   0.66666669f

Definition at line 71 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_SPEND_SLOPE_S

#define PSY_3GPP_SPEND_SLOPE_S   0.81818181f

Definition at line 72 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_SPEND_ADD_L

#define PSY_3GPP_SPEND_ADD_L   -0.35f

Definition at line 73 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_SPEND_ADD_S

#define PSY_3GPP_SPEND_ADD_S   -0.26111111f

Definition at line 74 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_CLIP_LO_L

#define PSY_3GPP_CLIP_LO_L   0.2f

Definition at line 75 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_CLIP_LO_S

#define PSY_3GPP_CLIP_LO_S   0.2f

Definition at line 76 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_CLIP_HI_L

#define PSY_3GPP_CLIP_HI_L   0.95f

Definition at line 77 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_CLIP_HI_S

#define PSY_3GPP_CLIP_HI_S   0.75f

Definition at line 78 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_AH_THR_LONG

#define PSY_3GPP_AH_THR_LONG   0.5f

Definition at line 80 of file aacpsy.c.

◆ PSY_3GPP_AH_THR_SHORT

#define PSY_3GPP_AH_THR_SHORT   0.63f

Definition at line 81 of file aacpsy.c.

◆ PSY_PE_FORGET_SLOPE

#define PSY_PE_FORGET_SLOPE   511

Definition at line 83 of file aacpsy.c.

Referenced by calc_bit_demand().

◆ PSY_3GPP_BITS_TO_PE

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

Definition at line 91 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ PSY_3GPP_PE_TO_BITS

#define PSY_3GPP_PE_TO_BITS (   bits)    ((bits) / 1.18f)

Definition at line 92 of file aacpsy.c.

◆ PSY_LAME_FIR_LEN

#define PSY_LAME_FIR_LEN   21

LAME psy model FIR order.

Definition at line 95 of file aacpsy.c.

Referenced by psy_hp_filter().

◆ AAC_BLOCK_SIZE_LONG

#define AAC_BLOCK_SIZE_LONG   1024

long block size

Definition at line 96 of file aacpsy.c.

Referenced by psy_3gpp_init(), and psy_hp_filter().

◆ AAC_BLOCK_SIZE_SHORT

#define AAC_BLOCK_SIZE_SHORT   128

short block size

Definition at line 97 of file aacpsy.c.

◆ AAC_NUM_BLOCKS_SHORT

#define AAC_NUM_BLOCKS_SHORT   8

number of blocks in a short sequence

Definition at line 98 of file aacpsy.c.

Referenced by lame_window_init(), and psy_lame_window().

◆ PSY_LAME_NUM_SUBBLOCKS

#define PSY_LAME_NUM_SUBBLOCKS   3

Number of sub-blocks in each short block.

Definition at line 99 of file aacpsy.c.

Referenced by lame_window_init().

◆ ATH_ADD

#define ATH_ADD   4

Definition at line 287 of file aacpsy.c.

Referenced by psy_3gpp_init().

Enumeration Type Documentation

◆ anonymous enum

anonymous enum
Enumerator
PSY_3GPP_AH_NONE 
PSY_3GPP_AH_INACTIVE 
PSY_3GPP_AH_ACTIVE 

Definition at line 85 of file aacpsy.c.

Function Documentation

◆ lame_calc_attack_threshold()

static float lame_calc_attack_threshold ( int  bitrate)
static

Calculate the ABR attack threshold from the above LAME psymodel table.

Definition at line 232 of file aacpsy.c.

Referenced by lame_window_init().

◆ lame_window_init()

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

LAME psy model specific initialization.

Definition at line 262 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ calc_bark()

static av_cold float calc_bark ( float  f)
static

Calculate Bark value for given line.

Definition at line 282 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ ath()

static av_cold float ath ( float  f,
float  add 
)
static

Calculate ATH value for given frequency.

Borrowed from Lame.

Definition at line 292 of file aacpsy.c.

Referenced by psy_3gpp_init().

◆ psy_3gpp_init()

static av_cold int psy_3gpp_init ( FFPsyContext ctx)
static

Definition at line 301 of file aacpsy.c.

◆ iir_filter()

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

IIR filter used in block switching decision.

Definition at line 384 of file aacpsy.c.

◆ psy_3gpp_window()

static av_unused 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 405 of file aacpsy.c.

◆ calc_bit_demand()

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

Definition at line 491 of file aacpsy.c.

◆ calc_pe_3gpp()

static float calc_pe_3gpp ( AacPsyBand band)
static

Definition at line 534 of file aacpsy.c.

◆ calc_reduction_3gpp()

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

Definition at line 557 of file aacpsy.c.

◆ calc_reduced_thr_3gpp()

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

Definition at line 571 of file aacpsy.c.

◆ calc_thr_3gpp()

static void calc_thr_3gpp ( const FFPsyWindowInfo wi,
const int  num_bands,
AacPsyChannel pch,
const uint8_t band_sizes,
const float *  coefs,
const int  cutoff 
)
static

Definition at line 597 of file aacpsy.c.

◆ psy_hp_filter()

static void psy_hp_filter ( const float *  firbuf,
float *  hpfsmpl,
const float *  psy_fir_coeffs 
)
static

Definition at line 628 of file aacpsy.c.

◆ psy_3gpp_analyze_channel()

static void psy_3gpp_analyze_channel ( FFPsyContext ctx,
int  channel,
const float *  coefs,
const FFPsyWindowInfo wi 
)
static

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

Definition at line 649 of file aacpsy.c.

Referenced by psy_3gpp_analyze().

◆ psy_3gpp_analyze()

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

Definition at line 845 of file aacpsy.c.

◆ psy_3gpp_end()

static av_cold void psy_3gpp_end ( FFPsyContext apc)
static

Definition at line 855 of file aacpsy.c.

◆ lame_apply_block_type()

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

Definition at line 862 of file aacpsy.c.

◆ psy_lame_window()

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

Definition at line 880 of file aacpsy.c.

Variable Documentation

◆ psy_abr_map

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 180 of file aacpsy.c.

◆ psy_vbr_map

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 201 of file aacpsy.c.

◆ psy_fir_coeffs

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 219 of file aacpsy.c.

◆ window_grouping

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 397 of file aacpsy.c.

◆ ff_aac_psy_model

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,
}
static av_cold void psy_3gpp_end(FFPsyContext *apc)
Definition: aacpsy.c:855
static FFPsyWindowInfo psy_lame_window(FFPsyContext *ctx, const float *audio, const float *la, int channel, int prev_type)
Definition: aacpsy.c:880
static void psy_3gpp_analyze(FFPsyContext *ctx, int channel, const float **coeffs, const FFPsyWindowInfo *wi)
Definition: aacpsy.c:845
static av_cold int psy_3gpp_init(FFPsyContext *ctx)
Definition: aacpsy.c:301

Definition at line 1018 of file aacpsy.c.

Referenced by ff_psy_init().