44 #define PSY_3GPP_THR_SPREAD_HI 1.5f // spreading factor for low-to-hi threshold spreading (15 dB/Bark)
45 #define PSY_3GPP_THR_SPREAD_LOW 3.0f // spreading factor for hi-to-low threshold spreading (30 dB/Bark)
47 #define PSY_3GPP_EN_SPREAD_HI_L1 2.0f
49 #define PSY_3GPP_EN_SPREAD_HI_L2 1.5f
51 #define PSY_3GPP_EN_SPREAD_HI_S 1.5f
53 #define PSY_3GPP_EN_SPREAD_LOW_L 3.0f
55 #define PSY_3GPP_EN_SPREAD_LOW_S 2.0f
57 #define PSY_3GPP_RPEMIN 0.01f
58 #define PSY_3GPP_RPELEV 2.0f
60 #define PSY_3GPP_C1 3.0f
61 #define PSY_3GPP_C2 1.3219281f
62 #define PSY_3GPP_C3 0.55935729f
64 #define PSY_SNR_1DB 7.9432821e-1f
65 #define PSY_SNR_25DB 3.1622776e-3f
67 #define PSY_3GPP_SAVE_SLOPE_L -0.46666667f
68 #define PSY_3GPP_SAVE_SLOPE_S -0.36363637f
69 #define PSY_3GPP_SAVE_ADD_L -0.84285712f
70 #define PSY_3GPP_SAVE_ADD_S -0.75f
71 #define PSY_3GPP_SPEND_SLOPE_L 0.66666669f
72 #define PSY_3GPP_SPEND_SLOPE_S 0.81818181f
73 #define PSY_3GPP_SPEND_ADD_L -0.35f
74 #define PSY_3GPP_SPEND_ADD_S -0.26111111f
75 #define PSY_3GPP_CLIP_LO_L 0.2f
76 #define PSY_3GPP_CLIP_LO_S 0.2f
77 #define PSY_3GPP_CLIP_HI_L 0.95f
78 #define PSY_3GPP_CLIP_HI_S 0.75f
80 #define PSY_3GPP_AH_THR_LONG 0.5f
81 #define PSY_3GPP_AH_THR_SHORT 0.63f
89 #define PSY_3GPP_BITS_TO_PE(bits) ((bits) * 1.18f)
92 #define PSY_LAME_FIR_LEN 21
93 #define AAC_BLOCK_SIZE_LONG 1024
94 #define AAC_BLOCK_SIZE_SHORT 128
95 #define AAC_NUM_BLOCKS_SHORT 8
96 #define PSY_LAME_NUM_SUBBLOCKS 3
216 -8.65163e-18 * 2, -0.00851586 * 2, -6.74764e-18 * 2, 0.0209036 * 2,
217 -3.36639e-17 * 2, -0.0438162 * 2, -1.54175e-17 * 2, 0.0931738 * 2,
218 -5.52212e-17 * 2, -0.313819 * 2
231 int lower_range = 12, upper_range = 12;
232 int lower_range_kbps = psy_abr_map[12].
quality;
233 int upper_range_kbps = psy_abr_map[12].
quality;
239 for (i = 1; i < 13; i++) {
240 if (
FFMAX(bitrate, psy_abr_map[i].quality) != bitrate) {
242 upper_range_kbps = psy_abr_map[i ].
quality;
244 lower_range_kbps = psy_abr_map[i - 1].
quality;
250 if ((upper_range_kbps - bitrate) > (bitrate - lower_range_kbps))
251 return psy_abr_map[lower_range].
st_lrm;
252 return psy_abr_map[upper_range].
st_lrm;
262 for (i = 0; i < avctx->
channels; i++) {
280 return 13.3f *
atanf(0.00076f * f) + 3.5f *
atanf((f / 7500.0f) * (f / 7500.0f));
291 return 3.64 * pow(f, -0.8)
292 - 6.8 * exp(-0.6 * (f - 3.4) * (f - 3.4))
293 + 6.0 * exp(-0.15 * (f - 8.7) * (f - 8.7))
294 + (0.6 + 0.04 * add) * 0.001 * f * f * f * f;
301 float prev, minscale, minath, minsnr, pe_min;
304 const float num_bark =
calc_bark((
float)bandwidth);
319 for (j = 0; j < 2; j++) {
323 float avg_chan_bits = chan_bitrate * (j ? 128.0f : 1024.0f) / ctx->
avctx->
sample_rate;
332 for (g = 0; g < ctx->
num_bands[j]; g++) {
334 bark =
calc_bark((i-1) * line_to_frequency);
335 coeffs[
g].
barks = (bark + prev) / 2.0;
338 for (g = 0; g < ctx->
num_bands[j] - 1; g++) {
340 float bark_width = coeffs[g+1].
barks - coeffs->
barks;
343 coeff->
spread_low[1] = pow(10.0, -bark_width * en_spread_low);
344 coeff->
spread_hi [1] = pow(10.0, -bark_width * en_spread_hi);
345 pe_min = bark_pe * bark_width;
346 minsnr =
exp2(pe_min / band_sizes[g]) - 1.5f;
350 for (g = 0; g < ctx->
num_bands[j]; g++) {
351 minscale =
ath(start * line_to_frequency,
ATH_ADD);
352 for (i = 1; i < band_sizes[
g]; i++)
353 minscale =
FFMIN(minscale,
ath((start + i) * line_to_frequency,
ATH_ADD));
354 coeffs[
g].
ath = minscale - minath;
355 start += band_sizes[
g];
377 ret = 0.7548f * (in - state[0]) + 0.5095f * state[1];
387 0xB6, 0x6C, 0xD8, 0xB2, 0x66, 0xC6, 0x96, 0x36, 0x36
395 const int16_t *audio,
397 int channel,
int prev_type)
401 int attack_ratio = br <= 16000 ? 18 : 10;
405 int next_type = pch->next_window_seq;
410 int switch_to_eight = 0;
411 float sum = 0.0, sum2 = 0.0;
414 for (i = 0; i < 8; i++) {
415 for (j = 0; j < 128; j++) {
422 for (i = 0; i < 8; i++) {
423 if (s[i] > pch->win_energy * attack_ratio) {
429 pch->win_energy = pch->win_energy*7/8 + sum2/64;
431 wi.window_type[1] = prev_type;
439 grouping = pch->next_grouping;
455 pch->next_window_seq = next_type;
457 for (i = 0; i < 3; i++)
458 wi.window_type[i] = prev_type;
469 for (i = 0; i < 8; i++) {
470 if (!((grouping >> i) & 1))
472 wi.grouping[lastgrp]++;
489 float clipped_pe, bit_save, bit_spend, bit_factor, fill_level;
493 fill_level = av_clipf((
float)ctx->
fill_level / size, clip_low, clip_high);
494 clipped_pe = av_clipf(pe, ctx->
pe.
min, ctx->
pe.
max);
495 bit_save = (fill_level + bitsave_add) * bitsave_slope;
496 assert(bit_save <= 0.3f && bit_save >= -0.05000001f);
497 bit_spend = (fill_level + bitspend_add) * bitspend_slope;
498 assert(bit_spend <= 0.5f && bit_spend >= -0.1f);
505 bit_factor = 1.0f - bit_save + ((bit_spend - bit_save) / (ctx->
pe.
max - ctx->
pe.
min)) * (clipped_pe - ctx->
pe.
min);
539 float thr_avg, reduction;
541 if(active_lines == 0.0)
544 thr_avg =
exp2f((a - pe) / (4.0f * active_lines));
545 reduction =
exp2f((a - desired_pe) / (4.0f * active_lines)) - thr_avg;
547 return FFMAX(reduction, 0.0f);
553 float thr = band->
thr;
557 thr = sqrtf(thr) + reduction;
575 #ifndef calc_thr_3gpp
577 const uint8_t *band_sizes,
const float *coefs)
582 for (g = 0; g < num_bands; g++) {
585 float form_factor = 0.0f;
588 for (i = 0; i < band_sizes[
g]; i++) {
589 band->
energy += coefs[start+i] * coefs[start+i];
590 form_factor += sqrtf(fabs(coefs[start+i]));
592 Temp = band->
energy > 0 ? sqrtf((
float)band_sizes[g] / band->
energy) : 0;
594 band->
nz_lines = form_factor * sqrtf(Temp);
596 start += band_sizes[
g];
602 #ifndef psy_hp_filter
611 sum1 += psy_fir_coeffs[j] * (firbuf[i + j] + firbuf[i +
PSY_LAME_FIR_LEN - j]);
612 sum2 += psy_fir_coeffs[j + 1] * (firbuf[i + j + 1] + firbuf[i +
PSY_LAME_FIR_LEN - j - 1]);
616 hpfsmpl[i] = (sum1 + sum2) * 32768.0f;
630 float desired_bits, desired_pe, delta_pe, reduction=
NAN, spread_en[128] = {0};
631 float a = 0.0f, active_lines = 0.0f, norm_fac = 0.0f;
632 float pe = pctx->chan_bitrate > 32000 ? 0.0f :
FFMAX(50.0f, 100.0f - pctx->chan_bitrate * 100.0f / 32000.0f);
633 const int num_bands = ctx->num_bands[wi->num_windows == 8];
634 const uint8_t *band_sizes = ctx->bands[wi->num_windows == 8];
642 for (w = 0; w < wi->num_windows*16; w += 16) {
646 spread_en[0] = bands[0].
energy;
647 for (
g = 1;
g < num_bands;
g++) {
648 bands[
g].
thr =
FFMAX(bands[
g].thr, bands[
g-1].thr * coeffs[
g].spread_hi[0]);
649 spread_en[w+
g] =
FFMAX(bands[
g].energy, spread_en[w+
g-1] * coeffs[
g].spread_hi[1]);
651 for (
g = num_bands - 2;
g >= 0;
g--) {
652 bands[
g].
thr =
FFMAX(bands[
g].thr, bands[
g+1].thr * coeffs[
g].spread_low[0]);
653 spread_en[w+
g] =
FFMAX(spread_en[w+
g], spread_en[w+
g+1] * coeffs[
g].spread_low[1]);
656 for (
g = 0;
g < num_bands;
g++) {
671 if (spread_en[w+
g] * avoid_hole_thr > band->
energy || coeffs[
g].
min_snr > 1.0f)
679 ctx->ch[channel].entropy = pe;
680 desired_bits =
calc_bit_demand(pctx, pe, ctx->bitres.bits, ctx->bitres.size, wi->num_windows == 8);
686 if (ctx->bitres.bits > 0)
691 if (desired_pe < pe) {
693 for (w = 0; w < wi->num_windows*16; w += 16) {
698 for (
g = 0;
g < num_bands;
g++) {
710 for (i = 0; i < 2; i++) {
711 float pe_no_ah = 0.0f, desired_pe_no_ah;
712 active_lines = a = 0.0f;
713 for (w = 0; w < wi->num_windows*16; w += 16) {
714 for (
g = 0;
g < num_bands;
g++) {
718 pe_no_ah += band->
pe;
724 desired_pe_no_ah =
FFMAX(desired_pe - (pe - pe_no_ah), 0.0f);
725 if (active_lines > 0.0f)
729 for (w = 0; w < wi->num_windows*16; w += 16) {
730 for (
g = 0;
g < num_bands;
g++) {
733 if (active_lines > 0.0f)
736 if (band->
thr > 0.0f)
743 delta_pe = desired_pe - pe;
744 if (fabs(delta_pe) > 0.05f * desired_pe)
748 if (pe < 1.15f * desired_pe) {
750 norm_fac = 1.0f / norm_fac;
751 for (w = 0; w < wi->num_windows*16; w += 16) {
752 for (
g = 0;
g < num_bands;
g++) {
756 float delta_sfb_pe = band->
norm_fac * norm_fac * delta_pe;
757 float thr = band->
thr;
769 while (pe > desired_pe &&
g--) {
770 for (w = 0; w < wi->num_windows*16; w+= 16) {
783 for (w = 0; w < wi->num_windows*16; w += 16) {
784 for (
g = 0;
g < num_bands;
g++) {
786 FFPsyBand *psy_band = &ctx->ch[channel].psy_bands[w+
g];
794 memcpy(pch->prev_band, pch->band,
sizeof(pch->band));
803 for (ch = 0; ch < group->
num_ch; ch++)
833 const float *la,
int channel,
int prev_type)
838 int uselongblock = 1;
846 float const *pf = hpfsmpl;
858 energy_subshort[i] = pch->prev_energy_subshort[i + ((
AAC_NUM_BLOCKS_SHORT - 1) * PSY_LAME_NUM_SUBBLOCKS)];
859 assert(pch->prev_energy_subshort[i + ((
AAC_NUM_BLOCKS_SHORT - 2) * PSY_LAME_NUM_SUBBLOCKS + 1)] > 0);
860 attack_intensity[i] = energy_subshort[i] / pch->prev_energy_subshort[i + ((
AAC_NUM_BLOCKS_SHORT - 2) * PSY_LAME_NUM_SUBBLOCKS + 1)];
861 energy_short[0] += energy_subshort[i];
867 for (; pf < pfe; pf++)
868 p =
FFMAX(p, fabsf(*pf));
878 if (p > energy_subshort[i + 1])
879 p = p / energy_subshort[i + 1];
880 else if (energy_subshort[i + 1] > p * 10.0f)
881 p = energy_subshort[i + 1] / (p * 10.0f);
889 if (!attacks[i / PSY_LAME_NUM_SUBBLOCKS])
890 if (attack_intensity[i] > pch->attack_threshold)
898 float const u = energy_short[i - 1];
899 float const v = energy_short[i];
900 float const m =
FFMAX(u, v);
902 if (u < 1.7f * v && v < 1.7f * u) {
903 if (i == 1 && attacks[0] < attacks[i])
908 att_sum += attacks[i];
911 if (attacks[0] <= pch->prev_attack)
914 att_sum += attacks[0];
916 if (pch->prev_attack == 3 || att_sum) {
919 for (i = 1; i < AAC_NUM_BLOCKS_SHORT + 1; i++)
920 if (attacks[i] && attacks[i-1])
937 max =
FFMAX(max, fabsf(wbuf[j]));
941 for (i = 0; i < 8; i++)
947 float clipping = 0.0f;
956 for (i = 0; i < 8; i++)
957 clipping =
FFMAX(clipping, clippings[i]);
964 for (i = 0; i < 8; i++) {
965 if (!((pch->next_grouping >> i) & 1))
970 for (i = 0; i < 8; i += wi.
grouping[i]) {
972 float clipping = 0.0f;
973 for (w = 0; w < wi.
grouping[i] && !clipping; w++)
974 clipping =
FFMAX(clipping, clippings[i+w]);
985 for (i = 0; i < 9; i++) {
993 pch->prev_attack = attacks[8];
1000 .
name =
"3GPP TS 26.403-inspired model",
int quality
Quality to map the rest of the vaules to.
static const uint8_t window_grouping[9]
window grouping information stored as bits (0 - new group, 1 - group continues)
int grouping[8]
window grouping (for e.g. AAC)
#define AAC_BLOCK_SIZE_SHORT
short block size
static int calc_bit_demand(AacPsyContext *ctx, float pe, int bits, int size, int short_window)
uint8_t ** bands
scalefactor band sizes for possible frame sizes
#define PSY_3GPP_AH_THR_SHORT
float iir_state[2]
hi-pass IIR filter state
static const PsyLamePreset psy_vbr_map[]
LAME psy model preset table for constant quality.
psychoacoustic information for an arbitrary group of channels
static float calc_reduction_3gpp(float a, float desired_pe, float pe, float active_lines)
float ath
absolute threshold of hearing per bands
#define PSY_3GPP_EN_SPREAD_HI_L1
static av_cold float ath(float f, float add)
Calculate ATH value for given frequency.
float prev_energy_subshort[AAC_NUM_BLOCKS_SHORT *PSY_LAME_NUM_SUBBLOCKS]
enum WindowSequence next_window_seq
window sequence to be used in the next frame
#define AAC_BLOCK_SIZE_LONG
long block size
struct FFPsyContext::@78 bitres
int * num_bands
number of scalefactor bands for possible frame sizes
Macro definitions for various function/variable attributes.
LAME psy model preset struct.
float thr
energy threshold
float correction
PE correction factor.
static av_cold void psy_3gpp_end(FFPsyContext *apc)
float attack_threshold
attack threshold for this channel
#define PSY_3GPP_EN_SPREAD_LOW_L
float nz_lines
number of non-zero spectral lines
psychoacoustic model frame type-dependent coefficients
int size
size of the bitresevoir in bits
static float calc_reduced_thr_3gpp(AacPsyBand *band, float min_snr, float reduction)
#define PSY_LAME_FIR_LEN
LAME psy model FIR order.
#define PSY_3GPP_CLIP_LO_L
#define PSY_3GPP_SPEND_SLOPE_S
#define PSY_3GPP_THR_SPREAD_LOW
context used by psychoacoustic model
single band psychoacoustic information
static float lame_calc_attack_threshold(int bitrate)
Calculate the ABR attack threshold from the above LAME psymodel table.
uint8_t next_grouping
stored grouping scheme for the next frame (in case of 8 short window sequence)
#define PSY_3GPP_SAVE_ADD_L
static void calc_thr_3gpp(const FFPsyWindowInfo *wi, const int num_bands, AacPsyChannel *pch, const uint8_t *band_sizes, const float *coefs)
static av_cold float calc_bark(float f)
Calculate Bark value for given line.
#define PSY_3GPP_SPEND_ADD_S
AacPsyBand prev_band[128]
bands information from the previous frame
3GPP TS26.403-inspired psychoacoustic model specific data
single/pair channel context for psychoacoustic model
static const float psy_fir_coeffs[]
LAME psy model FIR coefficient table.
float barks
Bark value for each spectral band in long frame.
int flags
AV_CODEC_FLAG_*.
float pe_const
constant part of the PE calculation
int num_windows
number of windows in a frame
static FFPsyWindowInfo psy_lame_window(FFPsyContext *ctx, const float *audio, const float *la, int channel, int prev_type)
#define PSY_3GPP_SPEND_SLOPE_L
#define PSY_3GPP_THR_SPREAD_HI
constants for 3GPP AAC psychoacoustic model
Libavcodec external API header.
codec-specific psychoacoustic model implementation
struct AacPsyContext::@31 pe
float thr_quiet
threshold in quiet
static void psy_3gpp_analyze(FFPsyContext *ctx, int channel, const float **coeffs, const FFPsyWindowInfo *wi)
int bit_rate
the average bitrate
#define AV_CODEC_FLAG_QSCALE
Use fixed qscale.
int prev_attack
attack value for the last short block in the previous sequence
#define PSY_3GPP_SAVE_SLOPE_S
uint8_t num_ch
number of channels in this group
int frame_bits
average bits per frame
int fill_level
bit reservoir fill level
static void lame_apply_block_type(AacPsyChannel *ctx, FFPsyWindowInfo *wi, int uselongblock)
#define PSY_3GPP_SAVE_SLOPE_L
Reference: libavcodec/aacpsy.c.
#define PSY_LAME_NUM_SUBBLOCKS
Number of sub-blocks in each short block.
const FFPsyModel ff_aac_psy_model
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.
float st_lrm
short threshold for L, R, and M channels
#define PSY_3GPP_EN_SPREAD_LOW_S
int sample_rate
samples per second
FFPsyChannelGroup * ff_psy_find_group(FFPsyContext *ctx, int channel)
Determine what group a channel belongs to.
main external API structure.
float win_energy
sliding average of channel energy
void * model_priv_data
psychoacoustic model implementation private data
float active_lines
number of active spectral lines
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(constuint8_t *) pi-0x80)*(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(constint16_t *) pi >>8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t,*(constint16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(constint32_t *) pi >>24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t,*(constint32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(constfloat *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(constfloat *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(constfloat *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(constdouble *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(constdouble *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(constdouble *) pi *(1U<< 31))))#defineSET_CONV_FUNC_GROUP(ofmt, ifmt) staticvoidset_generic_function(AudioConvert *ac){}voidff_audio_convert_free(AudioConvert **ac){if(!*ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);}AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enumAVSampleFormatout_fmt, enumAVSampleFormatin_fmt, intchannels, intsample_rate, intapply_map){AudioConvert *ac;intin_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) returnNULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method!=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt)>2){ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc){av_free(ac);returnNULL;}returnac;}in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar){ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar?ac->channels:1;}elseif(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;elseac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);returnac;}intff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in){intuse_generic=1;intlen=in->nb_samples;intp;if(ac->dc){av_log(ac->avr, AV_LOG_TRACE,"%dsamples-audio_convert:%sto%s(dithered)\n", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));returnff_convert_dither(ac-> in
static float iir_filter(int in, float state[2])
IIR filter used in block switching decision.
int avoid_holes
hole avoidance flag
Replacements for frequently missing libm functions.
AacPsyBand band[128]
bands information
#define PSY_3GPP_CLIP_HI_S
static const PsyLamePreset psy_abr_map[]
LAME psy model preset table for ABR.
int window_shape
window shape (sine/KBD/whatever)
float max
maximum allowed PE for bit factor calculation
float previous
allowed PE of the previous frame
AacPsyCoeffs psy_coef[2][64]
float min
minimum allowed PE for bit factor calculation
int global_quality
Global quality for codecs which cannot change it per frame.
static av_cold int psy_3gpp_init(FFPsyContext *ctx)
static void psy_hp_filter(const float *firbuf, float *hpfsmpl, const float *psy_fir_coeffs)
float spread_hi[2]
spreading factor for high-to-low threshold spreading in long frame
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.
static float calc_pe_3gpp(AacPsyBand *band)
windowing related information
#define PSY_3GPP_BITS_TO_PE(bits)
float norm_fac
normalization factor for linearization
int chan_bitrate
bitrate per channel
int cutoff
Audio cutoff bandwidth (0 means "automatic")
#define PSY_3GPP_CLIP_LO_S
#define PSY_3GPP_AH_THR_LONG
static const int16_t coeffs[]
int channels
number of audio channels
float pe
perceptual entropy
#define PSY_3GPP_EN_SPREAD_HI_S
static const double coeff[2][5]
#define FF_QP2LAMBDA
factor to convert from H.263 QP to lambda
static void * av_mallocz_array(size_t nmemb, size_t size)
#define PSY_3GPP_SAVE_ADD_S
information for single band used by 3GPP TS26.403-inspired psychoacoustic model
AVCodecContext * avctx
encoder context
float spread_low[2]
spreading factor for low-to-high threshold spreading in long frame
#define PSY_3GPP_CLIP_HI_L
int window_type[3]
window type (short/long/transitional, etc.) - current, previous and next
void * av_mallocz(size_t size)
Allocate a block of size bytes with alignment suitable for all memory accesses (including vectors if ...
#define AAC_NUM_BLOCKS_SHORT
number of blocks in a short sequence
float clipping[8]
maximum absolute normalized intensity in the given window for clip avoidance
#define PSY_3GPP_SPEND_ADD_L
static av_cold void lame_window_init(AacPsyContext *ctx, AVCodecContext *avctx)
LAME psy model specific initialization.