projects / libav.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
aac: Don't bias output during the IMDCT if SBR will be used.
[libav.git] / libavcodec / aacsbr.c
1 /*
2 * AAC Spectral Band Replication decoding functions
3 * Copyright (c) 2008-2009 Robert Swain ( rob opendot cl )
4 * Copyright (c) 2009-2010 Alex Converse <alex.converse@gmail.com>
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 /**
24 * @file libavcodec/aacsbr.c
25 * AAC Spectral Band Replication decoding functions
26 * @author Robert Swain ( rob opendot cl )
27 */
28
29 #include "aac.h"
30 #include "sbr.h"
31 #include "aacsbr.h"
32 #include "aacsbrdata.h"
33 #include "fft.h"
34
35 #include <stdint.h>
36 #include <float.h>
37
38 #define ENVELOPE_ADJUSTMENT_OFFSET 2
39 #define NOISE_FLOOR_OFFSET 6.0f
40
41 /**
42 * SBR VLC tables
43 */
44 enum {
45 T_HUFFMAN_ENV_1_5DB,
46 F_HUFFMAN_ENV_1_5DB,
47 T_HUFFMAN_ENV_BAL_1_5DB,
48 F_HUFFMAN_ENV_BAL_1_5DB,
49 T_HUFFMAN_ENV_3_0DB,
50 F_HUFFMAN_ENV_3_0DB,
51 T_HUFFMAN_ENV_BAL_3_0DB,
52 F_HUFFMAN_ENV_BAL_3_0DB,
53 T_HUFFMAN_NOISE_3_0DB,
54 T_HUFFMAN_NOISE_BAL_3_0DB,
55 };
56
57 /**
58 * bs_frame_class - frame class of current SBR frame (14496-3 sp04 p98)
59 */
60 enum {
61 FIXFIX,
62 FIXVAR,
63 VARFIX,
64 VARVAR,
65 };
66
67 enum {
68 EXTENSION_ID_PS = 2,
69 };
70
71 static VLC vlc_sbr[10];
72 static const int8_t vlc_sbr_lav[10] =
73 { 60, 60, 24, 24, 31, 31, 12, 12, 31, 12 };
74 static DECLARE_ALIGNED(16, float, analysis_cos_pre)[64];
75 static DECLARE_ALIGNED(16, float, analysis_sin_pre)[64];
76 static DECLARE_ALIGNED(16, float, analysis_cossin_post)[32][2];
77 static const DECLARE_ALIGNED(16, float, zero64)[64];
78
79 #define SBR_INIT_VLC_STATIC(num, size) \
80 INIT_VLC_STATIC(&vlc_sbr[num], 9, sbr_tmp[num].table_size / sbr_tmp[num].elem_size, \
81 sbr_tmp[num].sbr_bits , 1, 1, \
82 sbr_tmp[num].sbr_codes, sbr_tmp[num].elem_size, sbr_tmp[num].elem_size, \
83 size)
84
85 #define SBR_VLC_ROW(name) \
86 { name ## _codes, name ## _bits, sizeof(name ## _codes), sizeof(name ## _codes[0]) }
87
88 av_cold void ff_aac_sbr_init(void)
89 {
90 int n, k;
91 static const struct {
92 const void *sbr_codes, *sbr_bits;
93 const unsigned int table_size, elem_size;
94 } sbr_tmp[] = {
95 SBR_VLC_ROW(t_huffman_env_1_5dB),
96 SBR_VLC_ROW(f_huffman_env_1_5dB),
97 SBR_VLC_ROW(t_huffman_env_bal_1_5dB),
98 SBR_VLC_ROW(f_huffman_env_bal_1_5dB),
99 SBR_VLC_ROW(t_huffman_env_3_0dB),
100 SBR_VLC_ROW(f_huffman_env_3_0dB),
101 SBR_VLC_ROW(t_huffman_env_bal_3_0dB),
102 SBR_VLC_ROW(f_huffman_env_bal_3_0dB),
103 SBR_VLC_ROW(t_huffman_noise_3_0dB),
104 SBR_VLC_ROW(t_huffman_noise_bal_3_0dB),
105 };
106
107 // SBR VLC table initialization
108 SBR_INIT_VLC_STATIC(0, 1098);
109 SBR_INIT_VLC_STATIC(1, 1092);
110 SBR_INIT_VLC_STATIC(2, 768);
111 SBR_INIT_VLC_STATIC(3, 1026);
112 SBR_INIT_VLC_STATIC(4, 1058);
113 SBR_INIT_VLC_STATIC(5, 1052);
114 SBR_INIT_VLC_STATIC(6, 544);
115 SBR_INIT_VLC_STATIC(7, 544);
116 SBR_INIT_VLC_STATIC(8, 592);
117 SBR_INIT_VLC_STATIC(9, 512);
118
119 for (n = 0; n < 64; n++) {
120 float pre = M_PI * n / 64;
121 analysis_cos_pre[n] = cosf(pre);
122 analysis_sin_pre[n] = sinf(pre);
123 }
124 for (k = 0; k < 32; k++) {
125 float post = M_PI * (k + 0.5) / 128;
126 analysis_cossin_post[k][0] = 4.0 * cosf(post);
127 analysis_cossin_post[k][1] = -4.0 * sinf(post);
128 }
129 for (n = 1; n < 320; n++)
130 sbr_qmf_window_us[320 + n] = sbr_qmf_window_us[320 - n];
131 sbr_qmf_window_us[384] = -sbr_qmf_window_us[384];
132 sbr_qmf_window_us[512] = -sbr_qmf_window_us[512];
133
134 for (n = 0; n < 320; n++)
135 sbr_qmf_window_ds[n] = sbr_qmf_window_us[2*n];
136 }
137
138 av_cold void ff_aac_sbr_ctx_init(SpectralBandReplication *sbr)
139 {
140 sbr->kx[0] = sbr->kx[1] = 32; //Typo in spec, kx' inits to 32
141 sbr->data[0].e_a[1] = sbr->data[1].e_a[1] = -1;
142 sbr->data[0].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128);
143 sbr->data[1].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128);
144 ff_mdct_init(&sbr->mdct, 7, 1, 1.0/64);
145 ff_rdft_init(&sbr->rdft, 6, IDFT_R2C);
146 }
147
148 av_cold void ff_aac_sbr_ctx_close(SpectralBandReplication *sbr)
149 {
150 ff_mdct_end(&sbr->mdct);
151 ff_rdft_end(&sbr->rdft);
152 }
153
154 static int qsort_comparison_function_int16(const void *a, const void *b)
155 {
156 return *(const int16_t *)a - *(const int16_t *)b;
157 }
158
159 static inline int in_table_int16(const int16_t *table, int last_el, int16_t needle)
160 {
161 int i;
162 for (i = 0; i <= last_el; i++)
163 if (table[i] == needle)
164 return 1;
165 return 0;
166 }
167
168 /// Limiter Frequency Band Table (14496-3 sp04 p198)
169 static void sbr_make_f_tablelim(SpectralBandReplication *sbr)
170 {
171 int k;
172 if (sbr->bs_limiter_bands > 0) {
173 static const float bands_warped[3] = { 1.32715174233856803909f, //2^(0.49/1.2)
174 1.18509277094158210129f, //2^(0.49/2)
175 1.11987160404675912501f }; //2^(0.49/3)
176 const float lim_bands_per_octave_warped = bands_warped[sbr->bs_limiter_bands - 1];
177 int16_t patch_borders[5];
178 uint16_t *in = sbr->f_tablelim + 1, *out = sbr->f_tablelim;
179
180 patch_borders[0] = sbr->kx[1];
181 for (k = 1; k <= sbr->num_patches; k++)
182 patch_borders[k] = patch_borders[k-1] + sbr->patch_num_subbands[k-1];
183
184 memcpy(sbr->f_tablelim, sbr->f_tablelow,
185 (sbr->n[0] + 1) * sizeof(sbr->f_tablelow[0]));
186 if (sbr->num_patches > 1)
187 memcpy(sbr->f_tablelim + sbr->n[0] + 1, patch_borders + 1,
188 (sbr->num_patches - 1) * sizeof(patch_borders[0]));
189
190 qsort(sbr->f_tablelim, sbr->num_patches + sbr->n[0],
191 sizeof(sbr->f_tablelim[0]),
192 qsort_comparison_function_int16);
193
194 sbr->n_lim = sbr->n[0] + sbr->num_patches - 1;
195 while (out < sbr->f_tablelim + sbr->n_lim) {
196 if (*in >= *out * lim_bands_per_octave_warped) {
197 *++out = *in++;
198 } else if (*in == *out ||
199 !in_table_int16(patch_borders, sbr->num_patches, *in)) {
200 in++;
201 sbr->n_lim--;
202 } else if (!in_table_int16(patch_borders, sbr->num_patches, *out)) {
203 *out = *in++;
204 sbr->n_lim--;
205 } else {
206 *++out = *in++;
207 }
208 }
209 } else {
210 sbr->f_tablelim[0] = sbr->f_tablelow[0];
211 sbr->f_tablelim[1] = sbr->f_tablelow[sbr->n[0]];
212 sbr->n_lim = 1;
213 }
214 }
215
216 static unsigned int read_sbr_header(SpectralBandReplication *sbr, GetBitContext *gb)
217 {
218 unsigned int cnt = get_bits_count(gb);
219 uint8_t bs_header_extra_1;
220 uint8_t bs_header_extra_2;
221 int old_bs_limiter_bands = sbr->bs_limiter_bands;
222 SpectrumParameters old_spectrum_params;
223
224 sbr->start = 1;
225
226 // Save last spectrum parameters variables to compare to new ones
227 memcpy(&old_spectrum_params, &sbr->spectrum_params, sizeof(SpectrumParameters));
228
229 sbr->bs_amp_res_header = get_bits1(gb);
230 sbr->spectrum_params.bs_start_freq = get_bits(gb, 4);
231 sbr->spectrum_params.bs_stop_freq = get_bits(gb, 4);
232 sbr->spectrum_params.bs_xover_band = get_bits(gb, 3);
233 skip_bits(gb, 2); // bs_reserved
234
235 bs_header_extra_1 = get_bits1(gb);
236 bs_header_extra_2 = get_bits1(gb);
237
238 if (bs_header_extra_1) {
239 sbr->spectrum_params.bs_freq_scale = get_bits(gb, 2);
240 sbr->spectrum_params.bs_alter_scale = get_bits1(gb);
241 sbr->spectrum_params.bs_noise_bands = get_bits(gb, 2);
242 } else {
243 sbr->spectrum_params.bs_freq_scale = 2;
244 sbr->spectrum_params.bs_alter_scale = 1;
245 sbr->spectrum_params.bs_noise_bands = 2;
246 }
247
248 // Check if spectrum parameters changed
249 if (memcmp(&old_spectrum_params, &sbr->spectrum_params, sizeof(SpectrumParameters)))
250 sbr->reset = 1;
251
252 if (bs_header_extra_2) {
253 sbr->bs_limiter_bands = get_bits(gb, 2);
254 sbr->bs_limiter_gains = get_bits(gb, 2);
255 sbr->bs_interpol_freq = get_bits1(gb);
256 sbr->bs_smoothing_mode = get_bits1(gb);
257 } else {
258 sbr->bs_limiter_bands = 2;
259 sbr->bs_limiter_gains = 2;
260 sbr->bs_interpol_freq = 1;
261 sbr->bs_smoothing_mode = 1;
262 }
263
264 if (sbr->bs_limiter_bands != old_bs_limiter_bands && !sbr->reset)
265 sbr_make_f_tablelim(sbr);
266
267 return get_bits_count(gb) - cnt;
268 }
269
270 static int array_min_int16(const int16_t *array, int nel)
271 {
272 int i, min = array[0];
273 for (i = 1; i < nel; i++)
274 min = FFMIN(array[i], min);
275 return min;
276 }
277
278 static void make_bands(int16_t* bands, int start, int stop, int num_bands)
279 {
280 int k, previous, present;
281 float base, prod;
282
283 base = powf((float)stop / start, 1.0f / num_bands);
284 prod = start;
285 previous = start;
286
287 for (k = 0; k < num_bands-1; k++) {
288 prod *= base;
289 present = lrintf(prod);
290 bands[k] = present - previous;
291 previous = present;
292 }
293 bands[num_bands-1] = stop - previous;
294 }
295
296 static int check_n_master(AVCodecContext *avccontext, int n_master, int bs_xover_band)
297 {
298 // Requirements (14496-3 sp04 p205)
299 if (n_master <= 0) {
300 av_log(avccontext, AV_LOG_ERROR, "Invalid n_master: %d\n", n_master);
301 return -1;
302 }
303 if (bs_xover_band >= n_master) {
304 av_log(avccontext, AV_LOG_ERROR,
305 "Invalid bitstream, crossover band index beyond array bounds: %d\n",
306 bs_xover_band);
307 return -1;
308 }
309 return 0;
310 }
311
312 /// Master Frequency Band Table (14496-3 sp04 p194)
313 static int sbr_make_f_master(AACContext *ac, SpectralBandReplication *sbr,
314 SpectrumParameters *spectrum)
315 {
316 unsigned int temp, max_qmf_subbands;
317 unsigned int start_min, stop_min;
318 int k;
319 const int8_t *sbr_offset_ptr;
320 int16_t stop_dk[13];
321
322 if (sbr->sample_rate < 32000) {
323 temp = 3000;
324 } else if (sbr->sample_rate < 64000) {
325 temp = 4000;
326 } else
327 temp = 5000;
328
329 start_min = ((temp << 7) + (sbr->sample_rate >> 1)) / sbr->sample_rate;
330 stop_min = ((temp << 8) + (sbr->sample_rate >> 1)) / sbr->sample_rate;
331
332 switch (sbr->sample_rate) {
333 case 16000:
334 sbr_offset_ptr = sbr_offset[0];
335 break;
336 case 22050:
337 sbr_offset_ptr = sbr_offset[1];
338 break;
339 case 24000:
340 sbr_offset_ptr = sbr_offset[2];
341 break;
342 case 32000:
343 sbr_offset_ptr = sbr_offset[3];
344 break;
345 case 44100: case 48000: case 64000:
346 sbr_offset_ptr = sbr_offset[4];
347 break;
348 case 88200: case 96000: case 128000: case 176400: case 192000:
349 sbr_offset_ptr = sbr_offset[5];
350 break;
351 default:
352 av_log(ac->avccontext, AV_LOG_ERROR,
353 "Unsupported sample rate for SBR: %d\n", sbr->sample_rate);
354 return -1;
355 }
356
357 sbr->k[0] = start_min + sbr_offset_ptr[spectrum->bs_start_freq];
358
359 if (spectrum->bs_stop_freq < 14) {
360 sbr->k[2] = stop_min;
361 make_bands(stop_dk, stop_min, 64, 13);
362 qsort(stop_dk, 13, sizeof(stop_dk[0]), qsort_comparison_function_int16);
363 for (k = 0; k < spectrum->bs_stop_freq; k++)
364 sbr->k[2] += stop_dk[k];
365 } else if (spectrum->bs_stop_freq == 14) {
366 sbr->k[2] = 2*sbr->k[0];
367 } else if (spectrum->bs_stop_freq == 15) {
368 sbr->k[2] = 3*sbr->k[0];
369 } else {
370 av_log(ac->avccontext, AV_LOG_ERROR,
371 "Invalid bs_stop_freq: %d\n", spectrum->bs_stop_freq);
372 return -1;
373 }
374 sbr->k[2] = FFMIN(64, sbr->k[2]);
375
376 // Requirements (14496-3 sp04 p205)
377 if (sbr->sample_rate <= 32000) {
378 max_qmf_subbands = 48;
379 } else if (sbr->sample_rate == 44100) {
380 max_qmf_subbands = 35;
381 } else if (sbr->sample_rate >= 48000)
382 max_qmf_subbands = 32;
383
384 if (sbr->k[2] - sbr->k[0] > max_qmf_subbands) {
385 av_log(ac->avccontext, AV_LOG_ERROR,
386 "Invalid bitstream, too many QMF subbands: %d\n", sbr->k[2] - sbr->k[0]);
387 return -1;
388 }
389
390 if (!spectrum->bs_freq_scale) {
391 unsigned int dk;
392 int k2diff;
393
394 dk = spectrum->bs_alter_scale + 1;
395 sbr->n_master = ((sbr->k[2] - sbr->k[0] + (dk&2)) >> dk) << 1;
396 if (check_n_master(ac->avccontext, sbr->n_master, sbr->spectrum_params.bs_xover_band))
397 return -1;
398
399 for (k = 1; k <= sbr->n_master; k++)
400 sbr->f_master[k] = dk;
401
402 k2diff = sbr->k[2] - sbr->k[0] - sbr->n_master * dk;
403 if (k2diff < 0) {
404 sbr->f_master[1]--;
405 sbr->f_master[2]-= (k2diff < 1);
406 } else if (k2diff) {
407 sbr->f_master[sbr->n_master]++;
408 }
409
410 sbr->f_master[0] = sbr->k[0];
411 for (k = 1; k <= sbr->n_master; k++)
412 sbr->f_master[k] += sbr->f_master[k - 1];
413
414 } else {
415 int half_bands = 7 - spectrum->bs_freq_scale; // bs_freq_scale = {1,2,3}
416 int two_regions, num_bands_0;
417 int vdk0_max, vdk1_min;
418 int16_t vk0[49];
419
420 if (49 * sbr->k[2] > 110 * sbr->k[0]) {
421 two_regions = 1;
422 sbr->k[1] = 2 * sbr->k[0];
423 } else {
424 two_regions = 0;
425 sbr->k[1] = sbr->k[2];
426 }
427
428 num_bands_0 = lrintf(half_bands * log2f(sbr->k[1] / (float)sbr->k[0])) * 2;
429
430 if (num_bands_0 <= 0) { // Requirements (14496-3 sp04 p205)
431 av_log(ac->avccontext, AV_LOG_ERROR, "Invalid num_bands_0: %d\n", num_bands_0);
432 return -1;
433 }
434
435 vk0[0] = 0;
436
437 make_bands(vk0+1, sbr->k[0], sbr->k[1], num_bands_0);
438
439 qsort(vk0 + 1, num_bands_0, sizeof(vk0[1]), qsort_comparison_function_int16);
440 vdk0_max = vk0[num_bands_0];
441
442 vk0[0] = sbr->k[0];
443 for (k = 1; k <= num_bands_0; k++) {
444 if (vk0[k] <= 0) { // Requirements (14496-3 sp04 p205)
445 av_log(ac->avccontext, AV_LOG_ERROR, "Invalid vDk0[%d]: %d\n", k, vk0[k]);
446 return -1;
447 }
448 vk0[k] += vk0[k-1];
449 }
450
451 if (two_regions) {
452 int16_t vk1[49];
453 float invwarp = spectrum->bs_alter_scale ? 0.76923076923076923077f
454 : 1.0f; // bs_alter_scale = {0,1}
455 int num_bands_1 = lrintf(half_bands * invwarp *
456 log2f(sbr->k[2] / (float)sbr->k[1])) * 2;
457
458 make_bands(vk1+1, sbr->k[1], sbr->k[2], num_bands_1);
459
460 vdk1_min = array_min_int16(vk1 + 1, num_bands_1);
461
462 if (vdk1_min < vdk0_max) {
463 int change;
464 qsort(vk1 + 1, num_bands_1, sizeof(vk1[1]), qsort_comparison_function_int16);
465 change = FFMIN(vdk0_max - vk1[1], (vk1[num_bands_1] - vk1[1]) >> 1);
466 vk1[1] += change;
467 vk1[num_bands_1] -= change;
468 }
469
470 qsort(vk1 + 1, num_bands_1, sizeof(vk1[1]), qsort_comparison_function_int16);
471
472 vk1[0] = sbr->k[1];
473 for (k = 1; k <= num_bands_1; k++) {
474 if (vk1[k] <= 0) { // Requirements (14496-3 sp04 p205)
475 av_log(ac->avccontext, AV_LOG_ERROR, "Invalid vDk1[%d]: %d\n", k, vk1[k]);
476 return -1;
477 }
478 vk1[k] += vk1[k-1];
479 }
480
481 sbr->n_master = num_bands_0 + num_bands_1;
482 if (check_n_master(ac->avccontext, sbr->n_master, sbr->spectrum_params.bs_xover_band))
483 return -1;
484 memcpy(&sbr->f_master[0], vk0,
485 (num_bands_0 + 1) * sizeof(sbr->f_master[0]));
486 memcpy(&sbr->f_master[num_bands_0 + 1], vk1 + 1,
487 num_bands_1 * sizeof(sbr->f_master[0]));
488
489 } else {
490 sbr->n_master = num_bands_0;
491 if (check_n_master(ac->avccontext, sbr->n_master, sbr->spectrum_params.bs_xover_band))
492 return -1;
493 memcpy(sbr->f_master, vk0, (num_bands_0 + 1) * sizeof(sbr->f_master[0]));
494 }
495 }
496
497 return 0;
498 }
499
500 /// High Frequency Generation - Patch Construction (14496-3 sp04 p216 fig. 4.46)
501 static int sbr_hf_calc_npatches(AACContext *ac, SpectralBandReplication *sbr)
502 {
503 int i, k, sb = 0;
504 int msb = sbr->k[0];
505 int usb = sbr->kx[1];
506 int goal_sb = ((1000 << 11) + (sbr->sample_rate >> 1)) / sbr->sample_rate;
507
508 sbr->num_patches = 0;
509
510 if (goal_sb < sbr->kx[1] + sbr->m[1]) {
511 for (k = 0; sbr->f_master[k] < goal_sb; k++) ;
512 } else
513 k = sbr->n_master;
514
515 do {
516 int odd = 0;
517 for (i = k; i == k || sb > (sbr->k[0] - 1 + msb - odd); i--) {
518 sb = sbr->f_master[i];
519 odd = (sb + sbr->k[0]) & 1;
520 }
521
522 sbr->patch_num_subbands[sbr->num_patches] = FFMAX(sb - usb, 0);
523 sbr->patch_start_subband[sbr->num_patches] = sbr->k[0] - odd - sbr->patch_num_subbands[sbr->num_patches];
524
525 if (sbr->patch_num_subbands[sbr->num_patches] > 0) {
526 usb = sb;
527 msb = sb;
528 sbr->num_patches++;
529 } else
530 msb = sbr->kx[1];
531
532 if (sbr->f_master[k] - sb < 3)
533 k = sbr->n_master;
534 } while (sb != sbr->kx[1] + sbr->m[1]);
535
536 if (sbr->patch_num_subbands[sbr->num_patches-1] < 3 && sbr->num_patches > 1)
537 sbr->num_patches--;
538
539 // Requirements (14496-3 sp04 p205) sets the maximum number of patches to 5
540 // However the Coding Technologies decoder check uses 6 patches
541 if (sbr->num_patches > 6) {
542 av_log(ac->avccontext, AV_LOG_ERROR, "Too many patches: %d\n", sbr->num_patches);
543 return -1;
544 }
545
546 return 0;
547 }
548
549 /// Derived Frequency Band Tables (14496-3 sp04 p197)
550 static int sbr_make_f_derived(AACContext *ac, SpectralBandReplication *sbr)
551 {
552 int k, temp;
553
554 sbr->n[1] = sbr->n_master - sbr->spectrum_params.bs_xover_band;
555 sbr->n[0] = (sbr->n[1] + 1) >> 1;
556
557 memcpy(sbr->f_tablehigh, &sbr->f_master[sbr->spectrum_params.bs_xover_band],
558 (sbr->n[1] + 1) * sizeof(sbr->f_master[0]));
559 sbr->m[1] = sbr->f_tablehigh[sbr->n[1]] - sbr->f_tablehigh[0];
560 sbr->kx[1] = sbr->f_tablehigh[0];
561
562 // Requirements (14496-3 sp04 p205)
563 if (sbr->kx[1] + sbr->m[1] > 64) {
564 av_log(ac->avccontext, AV_LOG_ERROR,
565 "Stop frequency border too high: %d\n", sbr->kx[1] + sbr->m[1]);
566 return -1;
567 }
568 if (sbr->kx[1] > 32) {
569 av_log(ac->avccontext, AV_LOG_ERROR, "Start frequency border too high: %d\n", sbr->kx[1]);
570 return -1;
571 }
572
573 sbr->f_tablelow[0] = sbr->f_tablehigh[0];
574 temp = sbr->n[1] & 1;
575 for (k = 1; k <= sbr->n[0]; k++)
576 sbr->f_tablelow[k] = sbr->f_tablehigh[2 * k - temp];
577
578 sbr->n_q = FFMAX(1, lrintf(sbr->spectrum_params.bs_noise_bands *
579 log2f(sbr->k[2] / (float)sbr->kx[1]))); // 0 <= bs_noise_bands <= 3
580 if (sbr->n_q > 5) {
581 av_log(ac->avccontext, AV_LOG_ERROR, "Too many noise floor scale factors: %d\n", sbr->n_q);
582 return -1;
583 }
584
585 sbr->f_tablenoise[0] = sbr->f_tablelow[0];
586 temp = 0;
587 for (k = 1; k <= sbr->n_q; k++) {
588 temp += (sbr->n[0] - temp) / (sbr->n_q + 1 - k);
589 sbr->f_tablenoise[k] = sbr->f_tablelow[temp];
590 }
591
592 if (sbr_hf_calc_npatches(ac, sbr) < 0)
593 return -1;
594
595 sbr_make_f_tablelim(sbr);
596
597 sbr->data[0].f_indexnoise = 0;
598 sbr->data[1].f_indexnoise = 0;
599
600 return 0;
601 }
602
603 static av_always_inline void get_bits1_vector(GetBitContext *gb, uint8_t *vec,
604 int elements)
605 {
606 int i;
607 for (i = 0; i < elements; i++) {
608 vec[i] = get_bits1(gb);
609 }
610 }
611
612 /** ceil(log2(index+1)) */
613 static const int8_t ceil_log2[] = {
614 0, 1, 2, 2, 3, 3,
615 };
616
617 static int read_sbr_grid(AACContext *ac, SpectralBandReplication *sbr,
618 GetBitContext *gb, SBRData *ch_data)
619 {
620 int i;
621 unsigned bs_pointer = 0;
622 // frameLengthFlag ? 15 : 16; 960 sample length frames unsupported; this value is numTimeSlots
623 int abs_bord_trail = 16;
624 int num_rel_lead, num_rel_trail;
625 unsigned bs_num_env_old = ch_data->bs_num_env;
626
627 ch_data->bs_freq_res[0] = ch_data->bs_freq_res[ch_data->bs_num_env];
628 ch_data->bs_amp_res = sbr->bs_amp_res_header;
629 ch_data->t_env_num_env_old = ch_data->t_env[bs_num_env_old];
630
631 switch (ch_data->bs_frame_class = get_bits(gb, 2)) {
632 case FIXFIX:
633 ch_data->bs_num_env = 1 << get_bits(gb, 2);
634 num_rel_lead = ch_data->bs_num_env - 1;
635 if (ch_data->bs_num_env == 1)
636 ch_data->bs_amp_res = 0;
637
638 if (ch_data->bs_num_env > 4) {
639 av_log(ac->avccontext, AV_LOG_ERROR,
640 "Invalid bitstream, too many SBR envelopes in FIXFIX type SBR frame: %d\n",
641 ch_data->bs_num_env);
642 return -1;
643 }
644
645 ch_data->t_env[0] = 0;
646 ch_data->t_env[ch_data->bs_num_env] = abs_bord_trail;
647
648 abs_bord_trail = (abs_bord_trail + (ch_data->bs_num_env >> 1)) /
649 ch_data->bs_num_env;
650 for (i = 0; i < num_rel_lead; i++)
651 ch_data->t_env[i + 1] = ch_data->t_env[i] + abs_bord_trail;
652
653 ch_data->bs_freq_res[1] = get_bits1(gb);
654 for (i = 1; i < ch_data->bs_num_env; i++)
655 ch_data->bs_freq_res[i + 1] = ch_data->bs_freq_res[1];
656 break;
657 case FIXVAR:
658 abs_bord_trail += get_bits(gb, 2);
659 num_rel_trail = get_bits(gb, 2);
660 ch_data->bs_num_env = num_rel_trail + 1;
661 ch_data->t_env[0] = 0;
662 ch_data->t_env[ch_data->bs_num_env] = abs_bord_trail;
663
664 for (i = 0; i < num_rel_trail; i++)
665 ch_data->t_env[ch_data->bs_num_env - 1 - i] =
666 ch_data->t_env[ch_data->bs_num_env - i] - 2 * get_bits(gb, 2) - 2;
667
668 bs_pointer = get_bits(gb, ceil_log2[ch_data->bs_num_env]);
669
670 for (i = 0; i < ch_data->bs_num_env; i++)
671 ch_data->bs_freq_res[ch_data->bs_num_env - i] = get_bits1(gb);
672 break;
673 case VARFIX:
674 ch_data->t_env[0] = get_bits(gb, 2);
675 num_rel_lead = get_bits(gb, 2);
676 ch_data->bs_num_env = num_rel_lead + 1;
677 ch_data->t_env[ch_data->bs_num_env] = abs_bord_trail;
678
679 for (i = 0; i < num_rel_lead; i++)
680 ch_data->t_env[i + 1] = ch_data->t_env[i] + 2 * get_bits(gb, 2) + 2;
681
682 bs_pointer = get_bits(gb, ceil_log2[ch_data->bs_num_env]);
683
684 get_bits1_vector(gb, ch_data->bs_freq_res + 1, ch_data->bs_num_env);
685 break;
686 case VARVAR:
687 ch_data->t_env[0] = get_bits(gb, 2);
688 abs_bord_trail += get_bits(gb, 2);
689 num_rel_lead = get_bits(gb, 2);
690 num_rel_trail = get_bits(gb, 2);
691 ch_data->bs_num_env = num_rel_lead + num_rel_trail + 1;
692
693 if (ch_data->bs_num_env > 5) {
694 av_log(ac->avccontext, AV_LOG_ERROR,
695 "Invalid bitstream, too many SBR envelopes in VARVAR type SBR frame: %d\n",
696 ch_data->bs_num_env);
697 return -1;
698 }
699
700 ch_data->t_env[ch_data->bs_num_env] = abs_bord_trail;
701
702 for (i = 0; i < num_rel_lead; i++)
703 ch_data->t_env[i + 1] = ch_data->t_env[i] + 2 * get_bits(gb, 2) + 2;
704 for (i = 0; i < num_rel_trail; i++)
705 ch_data->t_env[ch_data->bs_num_env - 1 - i] =
706 ch_data->t_env[ch_data->bs_num_env - i] - 2 * get_bits(gb, 2) - 2;
707
708 bs_pointer = get_bits(gb, ceil_log2[ch_data->bs_num_env]);
709
710 get_bits1_vector(gb, ch_data->bs_freq_res + 1, ch_data->bs_num_env);
711 break;
712 }
713
714 if (bs_pointer > ch_data->bs_num_env + 1) {
715 av_log(ac->avccontext, AV_LOG_ERROR,
716 "Invalid bitstream, bs_pointer points to a middle noise border outside the time borders table: %d\n",
717 bs_pointer);
718 return -1;
719 }
720
721 ch_data->bs_num_noise = (ch_data->bs_num_env > 1) + 1;
722
723 ch_data->t_q[0] = ch_data->t_env[0];
724 ch_data->t_q[ch_data->bs_num_noise] = ch_data->t_env[ch_data->bs_num_env];
725 if (ch_data->bs_num_noise > 1) {
726 unsigned int idx;
727 if (ch_data->bs_frame_class == FIXFIX) {
728 idx = ch_data->bs_num_env >> 1;
729 } else if (ch_data->bs_frame_class & 1) { // FIXVAR or VARVAR
730 idx = ch_data->bs_num_env - FFMAX(bs_pointer - 1, 1);
731 } else { // VARFIX
732 if (!bs_pointer)
733 idx = 1;
734 else if (bs_pointer == 1)
735 idx = ch_data->bs_num_env - 1;
736 else // bs_pointer > 1
737 idx = bs_pointer - 1;
738 }
739 ch_data->t_q[1] = ch_data->t_env[idx];
740 }
741
742 ch_data->e_a[0] = -(ch_data->e_a[1] != bs_num_env_old); // l_APrev
743 ch_data->e_a[1] = -1;
744 if ((ch_data->bs_frame_class & 1) && bs_pointer) { // FIXVAR or VARVAR and bs_pointer != 0
745 ch_data->e_a[1] = ch_data->bs_num_env + 1 - bs_pointer;
746 } else if ((ch_data->bs_frame_class == 2) && (bs_pointer > 1)) // VARFIX and bs_pointer > 1
747 ch_data->e_a[1] = bs_pointer - 1;
748
749 return 0;
750 }
751
752 static void copy_sbr_grid(SBRData *dst, const SBRData *src) {
753 //These variables are saved from the previous frame rather than copied
754 dst->bs_freq_res[0] = dst->bs_freq_res[dst->bs_num_env];
755 dst->t_env_num_env_old = dst->t_env[dst->bs_num_env];
756 dst->e_a[0] = -(dst->e_a[1] != dst->bs_num_env);
757
758 //These variables are read from the bitstream and therefore copied
759 memcpy(dst->bs_freq_res+1, src->bs_freq_res+1, sizeof(dst->bs_freq_res)-sizeof(*dst->bs_freq_res));
760 memcpy(dst->t_env, src->t_env, sizeof(dst->t_env));
761 memcpy(dst->t_q, src->t_q, sizeof(dst->t_q));
762 dst->bs_num_env = src->bs_num_env;
763 dst->bs_amp_res = src->bs_amp_res;
764 dst->bs_num_noise = src->bs_num_noise;
765 dst->bs_frame_class = src->bs_frame_class;
766 dst->e_a[1] = src->e_a[1];
767 }
768
769 /// Read how the envelope and noise floor data is delta coded
770 static void read_sbr_dtdf(SpectralBandReplication *sbr, GetBitContext *gb,
771 SBRData *ch_data)
772 {
773 get_bits1_vector(gb, ch_data->bs_df_env, ch_data->bs_num_env);
774 get_bits1_vector(gb, ch_data->bs_df_noise, ch_data->bs_num_noise);
775 }
776
777 /// Read inverse filtering data
778 static void read_sbr_invf(SpectralBandReplication *sbr, GetBitContext *gb,
779 SBRData *ch_data)
780 {
781 int i;
782
783 memcpy(ch_data->bs_invf_mode[1], ch_data->bs_invf_mode[0], 5 * sizeof(uint8_t));
784 for (i = 0; i < sbr->n_q; i++)
785 ch_data->bs_invf_mode[0][i] = get_bits(gb, 2);
786 }
787
788 static void read_sbr_envelope(SpectralBandReplication *sbr, GetBitContext *gb,
789 SBRData *ch_data, int ch)
790 {
791 int bits;
792 int i, j, k;
793 VLC_TYPE (*t_huff)[2], (*f_huff)[2];
794 int t_lav, f_lav;
795 const int delta = (ch == 1 && sbr->bs_coupling == 1) + 1;
796 const int odd = sbr->n[1] & 1;
797
798 if (sbr->bs_coupling && ch) {
799 if (ch_data->bs_amp_res) {
800 bits = 5;
801 t_huff = vlc_sbr[T_HUFFMAN_ENV_BAL_3_0DB].table;
802 t_lav = vlc_sbr_lav[T_HUFFMAN_ENV_BAL_3_0DB];
803 f_huff = vlc_sbr[F_HUFFMAN_ENV_BAL_3_0DB].table;
804 f_lav = vlc_sbr_lav[F_HUFFMAN_ENV_BAL_3_0DB];
805 } else {
806 bits = 6;
807 t_huff = vlc_sbr[T_HUFFMAN_ENV_BAL_1_5DB].table;
808 t_lav = vlc_sbr_lav[T_HUFFMAN_ENV_BAL_1_5DB];
809 f_huff = vlc_sbr[F_HUFFMAN_ENV_BAL_1_5DB].table;
810 f_lav = vlc_sbr_lav[F_HUFFMAN_ENV_BAL_1_5DB];
811 }
812 } else {
813 if (ch_data->bs_amp_res) {
814 bits = 6;
815 t_huff = vlc_sbr[T_HUFFMAN_ENV_3_0DB].table;
816 t_lav = vlc_sbr_lav[T_HUFFMAN_ENV_3_0DB];
817 f_huff = vlc_sbr[F_HUFFMAN_ENV_3_0DB].table;
818 f_lav = vlc_sbr_lav[F_HUFFMAN_ENV_3_0DB];
819 } else {
820 bits = 7;
821 t_huff = vlc_sbr[T_HUFFMAN_ENV_1_5DB].table;
822 t_lav = vlc_sbr_lav[T_HUFFMAN_ENV_1_5DB];
823 f_huff = vlc_sbr[F_HUFFMAN_ENV_1_5DB].table;
824 f_lav = vlc_sbr_lav[F_HUFFMAN_ENV_1_5DB];
825 }
826 }
827
828 for (i = 0; i < ch_data->bs_num_env; i++) {
829 if (ch_data->bs_df_env[i]) {
830 // bs_freq_res[0] == bs_freq_res[bs_num_env] from prev frame
831 if (ch_data->bs_freq_res[i + 1] == ch_data->bs_freq_res[i]) {
832 for (j = 0; j < sbr->n[ch_data->bs_freq_res[i + 1]]; j++)
833 ch_data->env_facs[i + 1][j] = ch_data->env_facs[i][j] + delta * (get_vlc2(gb, t_huff, 9, 3) - t_lav);
834 } else if (ch_data->bs_freq_res[i + 1]) {
835 for (j = 0; j < sbr->n[ch_data->bs_freq_res[i + 1]]; j++) {
836 k = (j + odd) >> 1; // find k such that f_tablelow[k] <= f_tablehigh[j] < f_tablelow[k + 1]
837 ch_data->env_facs[i + 1][j] = ch_data->env_facs[i][k] + delta * (get_vlc2(gb, t_huff, 9, 3) - t_lav);
838 }
839 } else {
840 for (j = 0; j < sbr->n[ch_data->bs_freq_res[i + 1]]; j++) {
841 k = j ? 2*j - odd : 0; // find k such that f_tablehigh[k] == f_tablelow[j]
842 ch_data->env_facs[i + 1][j] = ch_data->env_facs[i][k] + delta * (get_vlc2(gb, t_huff, 9, 3) - t_lav);
843 }
844 }
845 } else {
846 ch_data->env_facs[i + 1][0] = delta * get_bits(gb, bits); // bs_env_start_value_balance
847 for (j = 1; j < sbr->n[ch_data->bs_freq_res[i + 1]]; j++)
848 ch_data->env_facs[i + 1][j] = ch_data->env_facs[i + 1][j - 1] + delta * (get_vlc2(gb, f_huff, 9, 3) - f_lav);
849 }
850 }
851
852 //assign 0th elements of env_facs from last elements
853 memcpy(ch_data->env_facs[0], ch_data->env_facs[ch_data->bs_num_env],
854 sizeof(ch_data->env_facs[0]));
855 }
856
857 static void read_sbr_noise(SpectralBandReplication *sbr, GetBitContext *gb,
858 SBRData *ch_data, int ch)
859 {
860 int i, j;
861 VLC_TYPE (*t_huff)[2], (*f_huff)[2];
862 int t_lav, f_lav;
863 int delta = (ch == 1 && sbr->bs_coupling == 1) + 1;
864
865 if (sbr->bs_coupling && ch) {
866 t_huff = vlc_sbr[T_HUFFMAN_NOISE_BAL_3_0DB].table;
867 t_lav = vlc_sbr_lav[T_HUFFMAN_NOISE_BAL_3_0DB];
868 f_huff = vlc_sbr[F_HUFFMAN_ENV_BAL_3_0DB].table;
869 f_lav = vlc_sbr_lav[F_HUFFMAN_ENV_BAL_3_0DB];
870 } else {
871 t_huff = vlc_sbr[T_HUFFMAN_NOISE_3_0DB].table;
872 t_lav = vlc_sbr_lav[T_HUFFMAN_NOISE_3_0DB];
873 f_huff = vlc_sbr[F_HUFFMAN_ENV_3_0DB].table;
874 f_lav = vlc_sbr_lav[F_HUFFMAN_ENV_3_0DB];
875 }
876
877 for (i = 0; i < ch_data->bs_num_noise; i++) {
878 if (ch_data->bs_df_noise[i]) {
879 for (j = 0; j < sbr->n_q; j++)
880 ch_data->noise_facs[i + 1][j] = ch_data->noise_facs[i][j] + delta * (get_vlc2(gb, t_huff, 9, 2) - t_lav);
881 } else {
882 ch_data->noise_facs[i + 1][0] = delta * get_bits(gb, 5); // bs_noise_start_value_balance or bs_noise_start_value_level
883 for (j = 1; j < sbr->n_q; j++)
884 ch_data->noise_facs[i + 1][j] = ch_data->noise_facs[i + 1][j - 1] + delta * (get_vlc2(gb, f_huff, 9, 3) - f_lav);
885 }
886 }
887
888 //assign 0th elements of noise_facs from last elements
889 memcpy(ch_data->noise_facs[0], ch_data->noise_facs[ch_data->bs_num_noise],
890 sizeof(ch_data->noise_facs[0]));
891 }
892
893 static void read_sbr_extension(AACContext *ac, SpectralBandReplication *sbr,
894 GetBitContext *gb,
895 int bs_extension_id, int *num_bits_left)
896 {
897 //TODO - implement ps_data for parametric stereo parsing
898 switch (bs_extension_id) {
899 case EXTENSION_ID_PS:
900 #if 0
901 *num_bits_left -= ff_ps_data(gb, ps);
902 #else
903 av_log_missing_feature(ac->avccontext, "Parametric Stereo is", 0);
904 skip_bits_long(gb, *num_bits_left); // bs_fill_bits
905 *num_bits_left = 0;
906 #endif
907 break;
908 default:
909 av_log_missing_feature(ac->avccontext, "Reserved SBR extensions are", 1);
910 skip_bits_long(gb, *num_bits_left); // bs_fill_bits
911 *num_bits_left = 0;
912 break;
913 }
914 }
915
916 static int read_sbr_single_channel_element(AACContext *ac,
917 SpectralBandReplication *sbr,
918 GetBitContext *gb)
919 {
920 if (get_bits1(gb)) // bs_data_extra
921 skip_bits(gb, 4); // bs_reserved
922
923 if (read_sbr_grid(ac, sbr, gb, &sbr->data[0]))
924 return -1;
925 read_sbr_dtdf(sbr, gb, &sbr->data[0]);
926 read_sbr_invf(sbr, gb, &sbr->data[0]);
927 read_sbr_envelope(sbr, gb, &sbr->data[0], 0);
928 read_sbr_noise(sbr, gb, &sbr->data[0], 0);
929
930 if ((sbr->data[0].bs_add_harmonic_flag = get_bits1(gb)))
931 get_bits1_vector(gb, sbr->data[0].bs_add_harmonic, sbr->n[1]);
932
933 return 0;
934 }
935
936 static int read_sbr_channel_pair_element(AACContext *ac,
937 SpectralBandReplication *sbr,
938 GetBitContext *gb)
939 {
940 if (get_bits1(gb)) // bs_data_extra
941 skip_bits(gb, 8); // bs_reserved
942
943 if ((sbr->bs_coupling = get_bits1(gb))) {
944 if (read_sbr_grid(ac, sbr, gb, &sbr->data[0]))
945 return -1;
946 copy_sbr_grid(&sbr->data[1], &sbr->data[0]);
947 read_sbr_dtdf(sbr, gb, &sbr->data[0]);
948 read_sbr_dtdf(sbr, gb, &sbr->data[1]);
949 read_sbr_invf(sbr, gb, &sbr->data[0]);
950 memcpy(sbr->data[1].bs_invf_mode[1], sbr->data[1].bs_invf_mode[0], sizeof(sbr->data[1].bs_invf_mode[0]));
951 memcpy(sbr->data[1].bs_invf_mode[0], sbr->data[0].bs_invf_mode[0], sizeof(sbr->data[1].bs_invf_mode[0]));
952 read_sbr_envelope(sbr, gb, &sbr->data[0], 0);
953 read_sbr_noise(sbr, gb, &sbr->data[0], 0);
954 read_sbr_envelope(sbr, gb, &sbr->data[1], 1);
955 read_sbr_noise(sbr, gb, &sbr->data[1], 1);
956 } else {
957 if (read_sbr_grid(ac, sbr, gb, &sbr->data[0]) ||
958 read_sbr_grid(ac, sbr, gb, &sbr->data[1]))
959 return -1;
960 read_sbr_dtdf(sbr, gb, &sbr->data[0]);
961 read_sbr_dtdf(sbr, gb, &sbr->data[1]);
962 read_sbr_invf(sbr, gb, &sbr->data[0]);
963 read_sbr_invf(sbr, gb, &sbr->data[1]);
964 read_sbr_envelope(sbr, gb, &sbr->data[0], 0);
965 read_sbr_envelope(sbr, gb, &sbr->data[1], 1);
966 read_sbr_noise(sbr, gb, &sbr->data[0], 0);
967 read_sbr_noise(sbr, gb, &sbr->data[1], 1);
968 }
969
970 if ((sbr->data[0].bs_add_harmonic_flag = get_bits1(gb)))
971 get_bits1_vector(gb, sbr->data[0].bs_add_harmonic, sbr->n[1]);
972 if ((sbr->data[1].bs_add_harmonic_flag = get_bits1(gb)))
973 get_bits1_vector(gb, sbr->data[1].bs_add_harmonic, sbr->n[1]);
974
975 return 0;
976 }
977
978 static unsigned int read_sbr_data(AACContext *ac, SpectralBandReplication *sbr,
979 GetBitContext *gb, int id_aac)
980 {
981 unsigned int cnt = get_bits_count(gb);
982
983 if (id_aac == TYPE_SCE || id_aac == TYPE_CCE) {
984 if (read_sbr_single_channel_element(ac, sbr, gb)) {
985 sbr->start = 0;
986 return get_bits_count(gb) - cnt;
987 }
988 } else if (id_aac == TYPE_CPE) {
989 if (read_sbr_channel_pair_element(ac, sbr, gb)) {
990 sbr->start = 0;
991 return get_bits_count(gb) - cnt;
992 }
993 } else {
994 av_log(ac->avccontext, AV_LOG_ERROR,
995 "Invalid bitstream - cannot apply SBR to element type %d\n", id_aac);
996 sbr->start = 0;
997 return get_bits_count(gb) - cnt;
998 }
999 if (get_bits1(gb)) { // bs_extended_data
1000 int num_bits_left = get_bits(gb, 4); // bs_extension_size
1001 if (num_bits_left == 15)
1002 num_bits_left += get_bits(gb, 8); // bs_esc_count
1003
1004 num_bits_left <<= 3;
1005 while (num_bits_left > 7) {
1006 num_bits_left -= 2;
1007 read_sbr_extension(ac, sbr, gb, get_bits(gb, 2), &num_bits_left); // bs_extension_id
1008 }
1009 }
1010
1011 return get_bits_count(gb) - cnt;
1012 }
1013
1014 static void sbr_reset(AACContext *ac, SpectralBandReplication *sbr)
1015 {
1016 int err;
1017 err = sbr_make_f_master(ac, sbr, &sbr->spectrum_params);
1018 if (err >= 0)
1019 err = sbr_make_f_derived(ac, sbr);
1020 if (err < 0) {
1021 av_log(ac->avccontext, AV_LOG_ERROR,
1022 "SBR reset failed. Switching SBR to pure upsampling mode.\n");
1023 sbr->start = 0;
1024 }
1025 }
1026
1027 /**
1028 * Decode Spectral Band Replication extension data; reference: table 4.55.
1029 *
1030 * @param crc flag indicating the presence of CRC checksum
1031 * @param cnt length of TYPE_FIL syntactic element in bytes
1032 *
1033 * @return Returns number of bytes consumed from the TYPE_FIL element.
1034 */
1035 int ff_decode_sbr_extension(AACContext *ac, SpectralBandReplication *sbr,
1036 GetBitContext *gb_host, int crc, int cnt, int id_aac)
1037 {
1038 unsigned int num_sbr_bits = 0, num_align_bits;
1039 unsigned bytes_read;
1040 GetBitContext gbc = *gb_host, *gb = &gbc;
1041 skip_bits_long(gb_host, cnt*8 - 4);
1042
1043 sbr->reset = 0;
1044
1045 if (!sbr->sample_rate)
1046 sbr->sample_rate = 2 * ac->m4ac.sample_rate; //TODO use the nominal sample rate for arbitrary sample rate support
1047 if (!ac->m4ac.ext_sample_rate)
1048 ac->m4ac.ext_sample_rate = 2 * ac->m4ac.sample_rate;
1049
1050 if (crc) {
1051 skip_bits(gb, 10); // bs_sbr_crc_bits; TODO - implement CRC check
1052 num_sbr_bits += 10;
1053 }
1054
1055 //Save some state from the previous frame.
1056 sbr->kx[0] = sbr->kx[1];
1057 sbr->m[0] = sbr->m[1];
1058
1059 num_sbr_bits++;
1060 if (get_bits1(gb)) // bs_header_flag
1061 num_sbr_bits += read_sbr_header(sbr, gb);
1062
1063 if (sbr->reset)
1064 sbr_reset(ac, sbr);
1065
1066 if (sbr->start)
1067 num_sbr_bits += read_sbr_data(ac, sbr, gb, id_aac);
1068
1069 num_align_bits = ((cnt << 3) - 4 - num_sbr_bits) & 7;
1070 bytes_read = ((num_sbr_bits + num_align_bits + 4) >> 3);
1071
1072 if (bytes_read > cnt) {
1073 av_log(ac->avccontext, AV_LOG_ERROR,
1074 "Expected to read %d SBR bytes actually read %d.\n", cnt, bytes_read);
1075 }
1076 return cnt;
1077 }
1078
1079 /// Dequantization and stereo decoding (14496-3 sp04 p203)
1080 static void sbr_dequant(SpectralBandReplication *sbr, int id_aac)
1081 {
1082 int k, e;
1083 int ch;
1084
1085 if (id_aac == TYPE_CPE && sbr->bs_coupling) {
1086 float alpha = sbr->data[0].bs_amp_res ? 1.0f : 0.5f;
1087 float pan_offset = sbr->data[0].bs_amp_res ? 12.0f : 24.0f;
1088 for (e = 1; e <= sbr->data[0].bs_num_env; e++) {
1089 for (k = 0; k < sbr->n[sbr->data[0].bs_freq_res[e]]; k++) {
1090 float temp1 = exp2f(sbr->data[0].env_facs[e][k] * alpha + 7.0f);
1091 float temp2 = exp2f((pan_offset - sbr->data[1].env_facs[e][k]) * alpha);
1092 float fac = temp1 / (1.0f + temp2);
1093 sbr->data[0].env_facs[e][k] = fac;
1094 sbr->data[1].env_facs[e][k] = fac * temp2;
1095 }
1096 }
1097 for (e = 1; e <= sbr->data[0].bs_num_noise; e++) {
1098 for (k = 0; k < sbr->n_q; k++) {
1099 float temp1 = exp2f(NOISE_FLOOR_OFFSET - sbr->data[0].noise_facs[e][k] + 1);
1100 float temp2 = exp2f(12 - sbr->data[1].noise_facs[e][k]);
1101 float fac = temp1 / (1.0f + temp2);
1102 sbr->data[0].noise_facs[e][k] = fac;
1103 sbr->data[1].noise_facs[e][k] = fac * temp2;
1104 }
1105 }
1106 } else { // SCE or one non-coupled CPE
1107 for (ch = 0; ch < (id_aac == TYPE_CPE) + 1; ch++) {
1108 float alpha = sbr->data[ch].bs_amp_res ? 1.0f : 0.5f;
1109 for (e = 1; e <= sbr->data[ch].bs_num_env; e++)
1110 for (k = 0; k < sbr->n[sbr->data[ch].bs_freq_res[e]]; k++)
1111 sbr->data[ch].env_facs[e][k] =
1112 exp2f(alpha * sbr->data[ch].env_facs[e][k] + 6.0f);
1113 for (e = 1; e <= sbr->data[ch].bs_num_noise; e++)
1114 for (k = 0; k < sbr->n_q; k++)
1115 sbr->data[ch].noise_facs[e][k] =
1116 exp2f(NOISE_FLOOR_OFFSET - sbr->data[ch].noise_facs[e][k]);
1117 }
1118 }
1119 }
1120
1121 /**
1122 * Analysis QMF Bank (14496-3 sp04 p206)
1123 *
1124 * @param x pointer to the beginning of the first sample window
1125 * @param W array of complex-valued samples split into subbands
1126 */
1127 static void sbr_qmf_analysis(DSPContext *dsp, RDFTContext *rdft, const float *in, float *x,
1128 float z[320], float W[2][32][32][2],
1129 float scale)
1130 {
1131 int i, k;
1132 memcpy(W[0], W[1], sizeof(W[0]));
1133 memcpy(x , x+1024, (320-32)*sizeof(x[0]));
1134 if (scale != 1.0f)
1135 for (i = 0; i < 1024; i++)
1136 x[288 + i] = in[i] * scale;
1137 else
1138 memcpy(x+288, in, 1024*sizeof(*x));
1139 for (i = 0; i < 32; i++) { // numTimeSlots*RATE = 16*2 as 960 sample frames
1140 // are not supported
1141 float re, im;
1142 dsp->vector_fmul_reverse(z, sbr_qmf_window_ds, x, 320);
1143 for (k = 0; k < 64; k++) {
1144 float f = z[k] + z[k + 64] + z[k + 128] + z[k + 192] + z[k + 256];
1145 z[k] = f * analysis_cos_pre[k];
1146 z[k+64] = f;
1147 }
1148 ff_rdft_calc(rdft, z);
1149 re = z[0] * 0.5f;
1150 im = 0.5f * dsp->scalarproduct_float(z+64, analysis_sin_pre, 64);
1151 W[1][i][0][0] = re * analysis_cossin_post[0][0] - im * analysis_cossin_post[0][1];
1152 W[1][i][0][1] = re * analysis_cossin_post[0][1] + im * analysis_cossin_post[0][0];
1153 for (k = 1; k < 32; k++) {
1154 re = z[2*k ] - re;
1155 im = z[2*k+1] - im;
1156 W[1][i][k][0] = re * analysis_cossin_post[k][0] - im * analysis_cossin_post[k][1];
1157 W[1][i][k][1] = re * analysis_cossin_post[k][1] + im * analysis_cossin_post[k][0];
1158 }
1159 x += 32;
1160 }
1161 }
1162
1163 /**
1164 * Synthesis QMF Bank (14496-3 sp04 p206) and Downsampled Synthesis QMF Bank
1165 * (14496-3 sp04 p206)
1166 */
1167 static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct,
1168 float *out, float X[2][32][64],
1169 float mdct_buf[2][64],
1170 float *v0, int *v_off, const unsigned int div,
1171 float bias, float scale)
1172 {
1173 int i, n;
1174 const float *sbr_qmf_window = div ? sbr_qmf_window_ds : sbr_qmf_window_us;
1175 int scale_and_bias = scale != 1.0f || bias != 0.0f;
1176 float *v;
1177 for (i = 0; i < 32; i++) {
1178 if (*v_off == 0) {
1179 int saved_samples = (1280 - 128) >> div;
1180 memcpy(&v0[SBR_SYNTHESIS_BUF_SIZE - saved_samples], v0, saved_samples * sizeof(float));
1181 *v_off = SBR_SYNTHESIS_BUF_SIZE - saved_samples - (128 >> div);
1182 } else {
1183 *v_off -= 128 >> div;
1184 }
1185 v = v0 + *v_off;
1186 for (n = 1; n < 64 >> div; n+=2) {
1187 X[1][i][n] = -X[1][i][n];
1188 }
1189 if (div) {
1190 memset(X[0][i]+32, 0, 32*sizeof(float));
1191 memset(X[1][i]+32, 0, 32*sizeof(float));
1192 }
1193 ff_imdct_half(mdct, mdct_buf[0], X[0][i]);
1194 ff_imdct_half(mdct, mdct_buf[1], X[1][i]);
1195 if (div) {
1196 for (n = 0; n < 32; n++) {
1197 v[ n] = -mdct_buf[0][63 - 2*n] + mdct_buf[1][2*n ];
1198 v[ 63 - n] = mdct_buf[0][62 - 2*n] + mdct_buf[1][2*n + 1];
1199 }
1200 } else {
1201 for (n = 0; n < 64; n++) {
1202 v[ n] = -mdct_buf[0][63 - n] + mdct_buf[1][ n ];
1203 v[127 - n] = mdct_buf[0][63 - n] + mdct_buf[1][ n ];
1204 }
1205 }
1206 dsp->vector_fmul_add(out, v , sbr_qmf_window , zero64, 64 >> div);
1207 dsp->vector_fmul_add(out, v + ( 192 >> div), sbr_qmf_window + ( 64 >> div), out , 64 >> div);
1208 dsp->vector_fmul_add(out, v + ( 256 >> div), sbr_qmf_window + (128 >> div), out , 64 >> div);
1209 dsp->vector_fmul_add(out, v + ( 448 >> div), sbr_qmf_window + (192 >> div), out , 64 >> div);
1210 dsp->vector_fmul_add(out, v + ( 512 >> div), sbr_qmf_window + (256 >> div), out , 64 >> div);
1211 dsp->vector_fmul_add(out, v + ( 704 >> div), sbr_qmf_window + (320 >> div), out , 64 >> div);
1212 dsp->vector_fmul_add(out, v + ( 768 >> div), sbr_qmf_window + (384 >> div), out , 64 >> div);
1213 dsp->vector_fmul_add(out, v + ( 960 >> div), sbr_qmf_window + (448 >> div), out , 64 >> div);
1214 dsp->vector_fmul_add(out, v + (1024 >> div), sbr_qmf_window + (512 >> div), out , 64 >> div);
1215 dsp->vector_fmul_add(out, v + (1216 >> div), sbr_qmf_window + (576 >> div), out , 64 >> div);
1216 if (scale_and_bias)
1217 for (n = 0; n < 64 >> div; n++)
1218 out[n] = out[n] * scale + bias;
1219 out += 64 >> div;
1220 }
1221 }
1222
1223 static void autocorrelate(const float x[40][2], float phi[3][2][2], int lag)
1224 {
1225 int i;
1226 float real_sum = 0.0f;
1227 float imag_sum = 0.0f;
1228 if (lag) {
1229 for (i = 1; i < 38; i++) {
1230 real_sum += x[i][0] * x[i+lag][0] + x[i][1] * x[i+lag][1];
1231 imag_sum += x[i][0] * x[i+lag][1] - x[i][1] * x[i+lag][0];
1232 }
1233 phi[2-lag][1][0] = real_sum + x[ 0][0] * x[lag][0] + x[ 0][1] * x[lag][1];
1234 phi[2-lag][1][1] = imag_sum + x[ 0][0] * x[lag][1] - x[ 0][1] * x[lag][0];
1235 if (lag == 1) {
1236 phi[0][0][0] = real_sum + x[38][0] * x[39][0] + x[38][1] * x[39][1];
1237 phi[0][0][1] = imag_sum + x[38][0] * x[39][1] - x[38][1] * x[39][0];
1238 }
1239 } else {
1240 for (i = 1; i < 38; i++) {
1241 real_sum += x[i][0] * x[i][0] + x[i][1] * x[i][1];
1242 }
1243 phi[2][1][0] = real_sum + x[ 0][0] * x[ 0][0] + x[ 0][1] * x[ 0][1];
1244 phi[1][0][0] = real_sum + x[38][0] * x[38][0] + x[38][1] * x[38][1];
1245 }
1246 }
1247
1248 /** High Frequency Generation (14496-3 sp04 p214+) and Inverse Filtering
1249 * (14496-3 sp04 p214)
1250 * Warning: This routine does not seem numerically stable.
1251 */
1252 static void sbr_hf_inverse_filter(float (*alpha0)[2], float (*alpha1)[2],
1253 const float X_low[32][40][2], int k0)
1254 {
1255 int k;
1256 for (k = 0; k < k0; k++) {
1257 float phi[3][2][2], dk;
1258
1259 autocorrelate(X_low[k], phi, 0);
1260 autocorrelate(X_low[k], phi, 1);
1261 autocorrelate(X_low[k], phi, 2);
1262
1263 dk = phi[2][1][0] * phi[1][0][0] -
1264 (phi[1][1][0] * phi[1][1][0] + phi[1][1][1] * phi[1][1][1]) / 1.000001f;
1265
1266 if (!dk) {
1267 alpha1[k][0] = 0;
1268 alpha1[k][1] = 0;
1269 } else {
1270 float temp_real, temp_im;
1271 temp_real = phi[0][0][0] * phi[1][1][0] -
1272 phi[0][0][1] * phi[1][1][1] -
1273 phi[0][1][0] * phi[1][0][0];
1274 temp_im = phi[0][0][0] * phi[1][1][1] +
1275 phi[0][0][1] * phi[1][1][0] -
1276 phi[0][1][1] * phi[1][0][0];
1277
1278 alpha1[k][0] = temp_real / dk;
1279 alpha1[k][1] = temp_im / dk;
1280 }
1281
1282 if (!phi[1][0][0]) {
1283 alpha0[k][0] = 0;
1284 alpha0[k][1] = 0;
1285 } else {
1286 float temp_real, temp_im;
1287 temp_real = phi[0][0][0] + alpha1[k][0] * phi[1][1][0] +
1288 alpha1[k][1] * phi[1][1][1];
1289 temp_im = phi[0][0][1] + alpha1[k][1] * phi[1][1][0] -
1290 alpha1[k][0] * phi[1][1][1];
1291
1292 alpha0[k][0] = -temp_real / phi[1][0][0];
1293 alpha0[k][1] = -temp_im / phi[1][0][0];
1294 }
1295
1296 if (alpha1[k][0] * alpha1[k][0] + alpha1[k][1] * alpha1[k][1] >= 16.0f ||
1297 alpha0[k][0] * alpha0[k][0] + alpha0[k][1] * alpha0[k][1] >= 16.0f) {
1298 alpha1[k][0] = 0;
1299 alpha1[k][1] = 0;
1300 alpha0[k][0] = 0;
1301 alpha0[k][1] = 0;
1302 }
1303 }
1304 }
1305
1306 /// Chirp Factors (14496-3 sp04 p214)
1307 static void sbr_chirp(SpectralBandReplication *sbr, SBRData *ch_data)
1308 {
1309 int i;
1310 float new_bw;
1311 static const float bw_tab[] = { 0.0f, 0.75f, 0.9f, 0.98f };
1312
1313 for (i = 0; i < sbr->n_q; i++) {
1314 if (ch_data->bs_invf_mode[0][i] + ch_data->bs_invf_mode[1][i] == 1) {
1315 new_bw = 0.6f;
1316 } else
1317 new_bw = bw_tab[ch_data->bs_invf_mode[0][i]];
1318
1319 if (new_bw < ch_data->bw_array[i]) {
1320 new_bw = 0.75f * new_bw + 0.25f * ch_data->bw_array[i];
1321 } else
1322 new_bw = 0.90625f * new_bw + 0.09375f * ch_data->bw_array[i];
1323 ch_data->bw_array[i] = new_bw < 0.015625f ? 0.0f : new_bw;
1324 }
1325 }
1326
1327 /// Generate the subband filtered lowband
1328 static int sbr_lf_gen(AACContext *ac, SpectralBandReplication *sbr,
1329 float X_low[32][40][2], const float W[2][32][32][2])
1330 {
1331 int i, k;
1332 const int t_HFGen = 8;
1333 const int i_f = 32;
1334 memset(X_low, 0, 32*sizeof(*X_low));
1335 for (k = 0; k < sbr->kx[1]; k++) {
1336 for (i = t_HFGen; i < i_f + t_HFGen; i++) {
1337 X_low[k][i][0] = W[1][i - t_HFGen][k][0];
1338 X_low[k][i][1] = W[1][i - t_HFGen][k][1];
1339 }
1340 }
1341 for (k = 0; k < sbr->kx[0]; k++) {
1342 for (i = 0; i < t_HFGen; i++) {
1343 X_low[k][i][0] = W[0][i + i_f - t_HFGen][k][0];
1344 X_low[k][i][1] = W[0][i + i_f - t_HFGen][k][1];
1345 }
1346 }
1347 return 0;
1348 }
1349
1350 /// High Frequency Generator (14496-3 sp04 p215)
1351 static int sbr_hf_gen(AACContext *ac, SpectralBandReplication *sbr,
1352 float X_high[64][40][2], const float X_low[32][40][2],
1353 const float (*alpha0)[2], const float (*alpha1)[2],
1354 const float bw_array[5], const uint8_t *t_env,
1355 int bs_num_env)
1356 {
1357 int i, j, x;
1358 int g = 0;
1359 int k = sbr->kx[1];
1360 for (j = 0; j < sbr->num_patches; j++) {
1361 for (x = 0; x < sbr->patch_num_subbands[j]; x++, k++) {
1362 float alpha[4];
1363 const int p = sbr->patch_start_subband[j] + x;
1364 while (g <= sbr->n_q && k >= sbr->f_tablenoise[g])
1365 g++;
1366 g--;
1367
1368 if (g < 0) {
1369 av_log(ac->avccontext, AV_LOG_ERROR,
1370 "ERROR : no subband found for frequency %d\n", k);
1371 return -1;
1372 }
1373
1374 alpha[0] = alpha1[p][0] * bw_array[g] * bw_array[g];
1375 alpha[1] = alpha1[p][1] * bw_array[g] * bw_array[g];
1376 alpha[2] = alpha0[p][0] * bw_array[g];
1377 alpha[3] = alpha0[p][1] * bw_array[g];
1378
1379 for (i = 2 * t_env[0]; i < 2 * t_env[bs_num_env]; i++) {
1380 const int idx = i + ENVELOPE_ADJUSTMENT_OFFSET;
1381 X_high[k][idx][0] =
1382 X_low[p][idx - 2][0] * alpha[0] -
1383 X_low[p][idx - 2][1] * alpha[1] +
1384 X_low[p][idx - 1][0] * alpha[2] -
1385 X_low[p][idx - 1][1] * alpha[3] +
1386 X_low[p][idx][0];
1387 X_high[k][idx][1] =
1388 X_low[p][idx - 2][1] * alpha[0] +
1389 X_low[p][idx - 2][0] * alpha[1] +
1390 X_low[p][idx - 1][1] * alpha[2] +
1391 X_low[p][idx - 1][0] * alpha[3] +
1392 X_low[p][idx][1];
1393 }
1394 }
1395 }
1396 if (k < sbr->m[1] + sbr->kx[1])
1397 memset(X_high + k, 0, (sbr->m[1] + sbr->kx[1] - k) * sizeof(*X_high));
1398
1399 return 0;
1400 }
1401
1402 /// Generate the subband filtered lowband
1403 static int sbr_x_gen(SpectralBandReplication *sbr, float X[2][32][64],
1404 const float X_low[32][40][2], const float Y[2][38][64][2],
1405 int ch)
1406 {
1407 int k, i;
1408 const int i_f = 32;
1409 const int i_Temp = FFMAX(2*sbr->data[ch].t_env_num_env_old - i_f, 0);
1410 memset(X, 0, 2*sizeof(*X));
1411 for (k = 0; k < sbr->kx[0]; k++) {
1412 for (i = 0; i < i_Temp; i++) {
1413 X[0][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][0];
1414 X[1][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][1];
1415 }
1416 }
1417 for (; k < sbr->kx[0] + sbr->m[0]; k++) {
1418 for (i = 0; i < i_Temp; i++) {
1419 X[0][i][k] = Y[0][i + i_f][k][0];
1420 X[1][i][k] = Y[0][i + i_f][k][1];
1421 }
1422 }
1423
1424 for (k = 0; k < sbr->kx[1]; k++) {
1425 for (i = i_Temp; i < i_f; i++) {
1426 X[0][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][0];
1427 X[1][i][k] = X_low[k][i + ENVELOPE_ADJUSTMENT_OFFSET][1];
1428 }
1429 }
1430 for (; k < sbr->kx[1] + sbr->m[1]; k++) {
1431 for (i = i_Temp; i < i_f; i++) {
1432 X[0][i][k] = Y[1][i][k][0];
1433 X[1][i][k] = Y[1][i][k][1];
1434 }
1435 }
1436 return 0;
1437 }
1438
1439 /** High Frequency Adjustment (14496-3 sp04 p217) and Mapping
1440 * (14496-3 sp04 p217)
1441 */
1442 static void sbr_mapping(AACContext *ac, SpectralBandReplication *sbr,
1443 SBRData *ch_data, int e_a[2])
1444 {
1445 int e, i, m;
1446
1447 memset(ch_data->s_indexmapped[1], 0, 7*sizeof(ch_data->s_indexmapped[1]));
1448 for (e = 0; e < ch_data->bs_num_env; e++) {
1449 const unsigned int ilim = sbr->n[ch_data->bs_freq_res[e + 1]];
1450 uint16_t *table = ch_data->bs_freq_res[e + 1] ? sbr->f_tablehigh : sbr->f_tablelow;
1451 int k;
1452
1453 for (i = 0; i < ilim; i++)
1454 for (m = table[i]; m < table[i + 1]; m++)
1455 sbr->e_origmapped[e][m - sbr->kx[1]] = ch_data->env_facs[e+1][i];
1456
1457 // ch_data->bs_num_noise > 1 => 2 noise floors
1458 k = (ch_data->bs_num_noise > 1) && (ch_data->t_env[e] >= ch_data->t_q[1]);
1459 for (i = 0; i < sbr->n_q; i++)
1460 for (m = sbr->f_tablenoise[i]; m < sbr->f_tablenoise[i + 1]; m++)
1461 sbr->q_mapped[e][m - sbr->kx[1]] = ch_data->noise_facs[k+1][i];
1462
1463 for (i = 0; i < sbr->n[1]; i++) {
1464 if (ch_data->bs_add_harmonic_flag) {
1465 const unsigned int m_midpoint =
1466 (sbr->f_tablehigh[i] + sbr->f_tablehigh[i + 1]) >> 1;
1467
1468 ch_data->s_indexmapped[e + 1][m_midpoint - sbr->kx[1]] = ch_data->bs_add_harmonic[i] *
1469 (e >= e_a[1] || (ch_data->s_indexmapped[0][m_midpoint - sbr->kx[1]] == 1));
1470 }
1471 }
1472
1473 for (i = 0; i < ilim; i++) {
1474 int additional_sinusoid_present = 0;
1475 for (m = table[i]; m < table[i + 1]; m++) {
1476 if (ch_data->s_indexmapped[e + 1][m - sbr->kx[1]]) {
1477 additional_sinusoid_present = 1;
1478 break;
1479 }
1480 }
1481 memset(&sbr->s_mapped[e][table[i] - sbr->kx[1]], additional_sinusoid_present,
1482 (table[i + 1] - table[i]) * sizeof(sbr->s_mapped[e][0]));
1483 }
1484 }
1485
1486 memcpy(ch_data->s_indexmapped[0], ch_data->s_indexmapped[ch_data->bs_num_env], sizeof(ch_data->s_indexmapped[0]));
1487 }
1488
1489 /// Estimation of current envelope (14496-3 sp04 p218)
1490 static void sbr_env_estimate(float (*e_curr)[48], float X_high[64][40][2],
1491 SpectralBandReplication *sbr, SBRData *ch_data)
1492 {
1493 int e, i, m;
1494
1495 if (sbr->bs_interpol_freq) {
1496 for (e = 0; e < ch_data->bs_num_env; e++) {
1497 const float recip_env_size = 0.5f / (ch_data->t_env[e + 1] - ch_data->t_env[e]);
1498 int ilb = ch_data->t_env[e] * 2 + ENVELOPE_ADJUSTMENT_OFFSET;
1499 int iub = ch_data->t_env[e + 1] * 2 + ENVELOPE_ADJUSTMENT_OFFSET;
1500
1501 for (m = 0; m < sbr->m[1]; m++) {
1502 float sum = 0.0f;
1503
1504 for (i = ilb; i < iub; i++) {
1505 sum += X_high[m + sbr->kx[1]][i][0] * X_high[m + sbr->kx[1]][i][0] +
1506 X_high[m + sbr->kx[1]][i][1] * X_high[m + sbr->kx[1]][i][1];
1507 }
1508 e_curr[e][m] = sum * recip_env_size;
1509 }
1510 }
1511 } else {
1512 int k, p;
1513
1514 for (e = 0; e < ch_data->bs_num_env; e++) {
1515 const int env_size = 2 * (ch_data->t_env[e + 1] - ch_data->t_env[e]);
1516 int ilb = ch_data->t_env[e] * 2 + ENVELOPE_ADJUSTMENT_OFFSET;
1517 int iub = ch_data->t_env[e + 1] * 2 + ENVELOPE_ADJUSTMENT_OFFSET;
1518 const uint16_t *table = ch_data->bs_freq_res[e + 1] ? sbr->f_tablehigh : sbr->f_tablelow;
1519
1520 for (p = 0; p < sbr->n[ch_data->bs_freq_res[e + 1]]; p++) {
1521 float sum = 0.0f;
1522 const int den = env_size * (table[p + 1] - table[p]);
1523
1524 for (k = table[p]; k < table[p + 1]; k++) {
1525 for (i = ilb; i < iub; i++) {
1526 sum += X_high[k][i][0] * X_high[k][i][0] +
1527 X_high[k][i][1] * X_high[k][i][1];
1528 }
1529 }
1530 sum /= den;
1531 for (k = table[p]; k < table[p + 1]; k++) {
1532 e_curr[e][k - sbr->kx[1]] = sum;
1533 }
1534 }
1535 }
1536 }
1537 }
1538
1539 /**
1540 * Calculation of levels of additional HF signal components (14496-3 sp04 p219)
1541 * and Calculation of gain (14496-3 sp04 p219)
1542 */
1543 static void sbr_gain_calc(AACContext *ac, SpectralBandReplication *sbr,
1544 SBRData *ch_data, const int e_a[2])
1545 {
1546 int e, k, m;
1547 // max gain limits : -3dB, 0dB, 3dB, inf dB (limiter off)
1548 static const float limgain[4] = { 0.70795, 1.0, 1.41254, 10000000000 };
1549
1550 for (e = 0; e < ch_data->bs_num_env; e++) {
1551 int delta = !((e == e_a[1]) || (e == e_a[0]));
1552 for (k = 0; k < sbr->n_lim; k++) {
1553 float gain_boost, gain_max;
1554 float sum[2] = { 0.0f, 0.0f };
1555 for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
1556 const float temp = sbr->e_origmapped[e][m] / (1.0f + sbr->q_mapped[e][m]);
1557 sbr->q_m[e][m] = sqrtf(temp * sbr->q_mapped[e][m]);
1558 sbr->s_m[e][m] = sqrtf(temp * ch_data->s_indexmapped[e + 1][m]);
1559 if (!sbr->s_mapped[e][m]) {
1560 sbr->gain[e][m] = sqrtf(sbr->e_origmapped[e][m] /
1561 ((1.0f + sbr->e_curr[e][m]) *
1562 (1.0f + sbr->q_mapped[e][m] * delta)));
1563 } else {
1564 sbr->gain[e][m] = sqrtf(sbr->e_origmapped[e][m] * sbr->q_mapped[e][m] /
1565 ((1.0f + sbr->e_curr[e][m]) *
1566 (1.0f + sbr->q_mapped[e][m])));
1567 }
1568 }
1569 for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
1570 sum[0] += sbr->e_origmapped[e][m];
1571 sum[1] += sbr->e_curr[e][m];
1572 }
1573 gain_max = limgain[sbr->bs_limiter_gains] * sqrtf((FLT_EPSILON + sum[0]) / (FLT_EPSILON + sum[1]));
1574 gain_max = FFMIN(100000, gain_max);
1575 for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
1576 float q_m_max = sbr->q_m[e][m] * gain_max / sbr->gain[e][m];
1577 sbr->q_m[e][m] = FFMIN(sbr->q_m[e][m], q_m_max);
1578 sbr->gain[e][m] = FFMIN(sbr->gain[e][m], gain_max);
1579 }
1580 sum[0] = sum[1] = 0.0f;
1581 for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
1582 sum[0] += sbr->e_origmapped[e][m];
1583 sum[1] += sbr->e_curr[e][m] * sbr->gain[e][m] * sbr->gain[e][m]
1584 + sbr->s_m[e][m] * sbr->s_m[e][m]
1585 + (delta && !sbr->s_m[e][m]) * sbr->q_m[e][m] * sbr->q_m[e][m];
1586 }
1587 gain_boost = sqrtf((FLT_EPSILON + sum[0]) / (FLT_EPSILON + sum[1]));
1588 gain_boost = FFMIN(1.584893192, gain_boost);
1589 for (m = sbr->f_tablelim[k] - sbr->kx[1]; m < sbr->f_tablelim[k + 1] - sbr->kx[1]; m++) {
1590 sbr->gain[e][m] *= gain_boost;
1591 sbr->q_m[e][m] *= gain_boost;
1592 sbr->s_m[e][m] *= gain_boost;
1593 }
1594 }
1595 }
1596 }
1597
1598 /// Assembling HF Signals (14496-3 sp04 p220)
1599 static void sbr_hf_assemble(float Y[2][38][64][2], const float X_high[64][40][2],
1600 SpectralBandReplication *sbr, SBRData *ch_data,
1601 const int e_a[2])
1602 {
1603 int e, i, j, m;
1604 const int h_SL = 4 * !sbr->bs_smoothing_mode;
1605 const int kx = sbr->kx[1];
1606 const int m_max = sbr->m[1];
1607 static const float h_smooth[5] = {
1608 0.33333333333333,
1609 0.30150283239582,
1610 0.21816949906249,
1611 0.11516383427084,
1612 0.03183050093751,
1613 };
1614 static const int8_t phi[2][4] = {
1615 { 1, 0, -1, 0}, // real
1616 { 0, 1, 0, -1}, // imaginary
1617 };
1618 float (*g_temp)[48] = ch_data->g_temp, (*q_temp)[48] = ch_data->q_temp;
1619 int indexnoise = ch_data->f_indexnoise;
1620 int indexsine = ch_data->f_indexsine;
1621 memcpy(Y[0], Y[1], sizeof(Y[0]));
1622
1623 if (sbr->reset) {
1624 for (i = 0; i < h_SL; i++) {
1625 memcpy(g_temp[i + 2*ch_data->t_env[0]], sbr->gain[0], m_max * sizeof(sbr->gain[0][0]));
1626 memcpy(q_temp[i + 2*ch_data->t_env[0]], sbr->q_m[0], m_max * sizeof(sbr->q_m[0][0]));
1627 }
1628 } else if (h_SL) {
1629 memcpy(g_temp[2*ch_data->t_env[0]], g_temp[2*ch_data->t_env_num_env_old], 4*sizeof(g_temp[0]));
1630 memcpy(q_temp[2*ch_data->t_env[0]], q_temp[2*ch_data->t_env_num_env_old], 4*sizeof(q_temp[0]));
1631 }
1632
1633 for (e = 0; e < ch_data->bs_num_env; e++) {
1634 for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) {
1635 memcpy(g_temp[h_SL + i], sbr->gain[e], m_max * sizeof(sbr->gain[0][0]));
1636 memcpy(q_temp[h_SL + i], sbr->q_m[e], m_max * sizeof(sbr->q_m[0][0]));
1637 }
1638 }
1639
1640 for (e = 0; e < ch_data->bs_num_env; e++) {
1641 for (i = 2 * ch_data->t_env[e]; i < 2 * ch_data->t_env[e + 1]; i++) {
1642 int phi_sign = (1 - 2*(kx & 1));
1643
1644 if (h_SL && e != e_a[0] && e != e_a[1]) {
1645 for (m = 0; m < m_max; m++) {
1646 const int idx1 = i + h_SL;
1647 float g_filt = 0.0f;
1648 for (j = 0; j <= h_SL; j++)
1649 g_filt += g_temp[idx1 - j][m] * h_smooth[j];
1650 Y[1][i][m + kx][0] =
1651 X_high[m + kx][i + ENVELOPE_ADJUSTMENT_OFFSET][0] * g_filt;
1652 Y[1][i][m + kx][1] =
1653 X_high[m + kx][i + ENVELOPE_ADJUSTMENT_OFFSET][1] * g_filt;
1654 }
1655 } else {
1656 for (m = 0; m < m_max; m++) {
1657 const float g_filt = g_temp[i + h_SL][m];
1658 Y[1][i][m + kx][0] =
1659 X_high[m + kx][i + ENVELOPE_ADJUSTMENT_OFFSET][0] * g_filt;
1660 Y[1][i][m + kx][1] =
1661 X_high[m + kx][i + ENVELOPE_ADJUSTMENT_OFFSET][1] * g_filt;
1662 }
1663 }
1664
1665 if (e != e_a[0] && e != e_a[1]) {
1666 for (m = 0; m < m_max; m++) {
1667 indexnoise = (indexnoise + 1) & 0x1ff;
1668 if (sbr->s_m[e][m]) {
1669 Y[1][i][m + kx][0] +=
1670 sbr->s_m[e][m] * phi[0][indexsine];
1671 Y[1][i][m + kx][1] +=
1672 sbr->s_m[e][m] * (phi[1][indexsine] * phi_sign);
1673 } else {
1674 float q_filt;
1675 if (h_SL) {
1676 const int idx1 = i + h_SL;
1677 q_filt = 0.0f;
1678 for (j = 0; j <= h_SL; j++)
1679 q_filt += q_temp[idx1 - j][m] * h_smooth[j];
1680 } else {
1681 q_filt = q_temp[i][m];
1682 }
1683 Y[1][i][m + kx][0] +=
1684 q_filt * sbr_noise_table[indexnoise][0];
1685 Y[1][i][m + kx][1] +=
1686 q_filt * sbr_noise_table[indexnoise][1];
1687 }
1688 phi_sign = -phi_sign;
1689 }
1690 } else {
1691 indexnoise = (indexnoise + m_max) & 0x1ff;
1692 for (m = 0; m < m_max; m++) {
1693 Y[1][i][m + kx][0] +=
1694 sbr->s_m[e][m] * phi[0][indexsine];
1695 Y[1][i][m + kx][1] +=
1696 sbr->s_m[e][m] * (phi[1][indexsine] * phi_sign);
1697 phi_sign = -phi_sign;
1698 }
1699 }
1700 indexsine = (indexsine + 1) & 3;
1701 }
1702 }
1703 ch_data->f_indexnoise = indexnoise;
1704 ch_data->f_indexsine = indexsine;
1705 }
1706
1707 void ff_sbr_dequant(AACContext *ac, SpectralBandReplication *sbr, int id_aac)
1708 {
1709 if (sbr->start) {
1710 sbr_dequant(sbr, id_aac);
1711 }
1712 }
1713
1714 void ff_sbr_apply(AACContext *ac, SpectralBandReplication *sbr, int ch,
1715 const float* in, float* out)
1716 {
1717 int downsampled = ac->m4ac.ext_sample_rate < sbr->sample_rate;
1718
1719 /* decode channel */
1720 sbr_qmf_analysis(&ac->dsp, &sbr->rdft, in, sbr->data[ch].analysis_filterbank_samples,
1721 (float*)sbr->qmf_filter_scratch,
1722 sbr->data[ch].W, 1/(-1024 * ac->sf_scale));
1723 sbr_lf_gen(ac, sbr, sbr->X_low, sbr->data[ch].W);
1724 if (sbr->start) {
1725 sbr_hf_inverse_filter(sbr->alpha0, sbr->alpha1, sbr->X_low, sbr->k[0]);
1726 sbr_chirp(sbr, &sbr->data[ch]);
1727 sbr_hf_gen(ac, sbr, sbr->X_high, sbr->X_low, sbr->alpha0, sbr->alpha1,
1728 sbr->data[ch].bw_array, sbr->data[ch].t_env,
1729 sbr->data[ch].bs_num_env);
1730
1731 // hf_adj
1732 sbr_mapping(ac, sbr, &sbr->data[ch], sbr->data[ch].e_a);
1733 sbr_env_estimate(sbr->e_curr, sbr->X_high, sbr, &sbr->data[ch]);
1734 sbr_gain_calc(ac, sbr, &sbr->data[ch], sbr->data[ch].e_a);
1735 sbr_hf_assemble(sbr->data[ch].Y, sbr->X_high, sbr, &sbr->data[ch],
1736 sbr->data[ch].e_a);
1737 }
1738
1739 /* synthesis */
1740 sbr_x_gen(sbr, sbr->X, sbr->X_low, sbr->data[ch].Y, ch);
1741 sbr_qmf_synthesis(&ac->dsp, &sbr->mdct, out, sbr->X, sbr->qmf_filter_scratch,
1742 sbr->data[ch].synthesis_filterbank_samples,
1743 &sbr->data[ch].synthesis_filterbank_samples_offset,
1744 downsampled,
1745 ac->add_bias, -1024 * ac->sf_scale);
1746 }
Libav master