FFmpeg
dss_sp.c
Go to the documentation of this file.
1 /*
2  * Digital Speech Standard - Standard Play mode (DSS SP) audio decoder.
3  * Copyright (C) 2014 Oleksij Rempel <linux@rempel-privat.de>
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 #include "libavutil/channel_layout.h"
23 #include "libavutil/common.h"
24 #include "libavutil/mem.h"
25 #include "libavutil/mem_internal.h"
26 #include "libavutil/opt.h"
27 
28 #include "avcodec.h"
29 #include "get_bits.h"
30 #include "internal.h"
31 
32 #define SUBFRAMES 4
33 #define PULSE_MAX 8
34 
35 #define DSS_SP_FRAME_SIZE 42
36 #define DSS_SP_SAMPLE_COUNT (66 * SUBFRAMES)
37 #define DSS_SP_FORMULA(a, b, c) ((int)((((a) * (1 << 15)) + (b) * (unsigned)(c)) + 0x4000) >> 15)
38 
39 typedef struct DssSpSubframe {
40  int16_t gain;
41  int32_t combined_pulse_pos;
42  int16_t pulse_pos[7];
43  int16_t pulse_val[7];
44 } DssSpSubframe;
45 
46 typedef struct DssSpFrame {
47  int16_t filter_idx[14];
48  int16_t sf_adaptive_gain[SUBFRAMES];
49  int16_t pitch_lag[SUBFRAMES];
50  struct DssSpSubframe sf[SUBFRAMES];
51 } DssSpFrame;
52 
53 typedef struct DssSpContext {
54  AVCodecContext *avctx;
55  int32_t excitation[288 + 6];
56  int32_t history[187];
57  DssSpFrame fparam;
58  int32_t working_buffer[SUBFRAMES][72];
59  int32_t audio_buf[15];
60  int32_t err_buf1[15];
61  int32_t lpc_filter[14];
62  int32_t filter[15];
63  int32_t vector_buf[72];
64  int noise_state;
65  int32_t err_buf2[15];
66 
67  int pulse_dec_mode;
68 
69  DECLARE_ALIGNED(16, uint8_t, bits)[DSS_SP_FRAME_SIZE +
70  AV_INPUT_BUFFER_PADDING_SIZE];
71 } DssSpContext;
72 
73 /*
74  * Used for the coding/decoding of the pulse positions for the MP-MLQ codebook.
75  */
76 static const uint32_t dss_sp_combinatorial_table[PULSE_MAX][72] = {
77  { 0, 0, 0, 0, 0, 0,
78  0, 0, 0, 0, 0, 0,
79  0, 0, 0, 0, 0, 0,
80  0, 0, 0, 0, 0, 0,
81  0, 0, 0, 0, 0, 0,
82  0, 0, 0, 0, 0, 0,
83  0, 0, 0, 0, 0, 0,
84  0, 0, 0, 0, 0, 0,
85  0, 0, 0, 0, 0, 0,
86  0, 0, 0, 0, 0, 0,
87  0, 0, 0, 0, 0, 0,
88  0, 0, 0, 0, 0, 0 },
89  { 0, 1, 2, 3, 4, 5,
90  6, 7, 8, 9, 10, 11,
91  12, 13, 14, 15, 16, 17,
92  18, 19, 20, 21, 22, 23,
93  24, 25, 26, 27, 28, 29,
94  30, 31, 32, 33, 34, 35,
95  36, 37, 38, 39, 40, 41,
96  42, 43, 44, 45, 46, 47,
97  48, 49, 50, 51, 52, 53,
98  54, 55, 56, 57, 58, 59,
99  60, 61, 62, 63, 64, 65,
100  66, 67, 68, 69, 70, 71 },
101  { 0, 0, 1, 3, 6, 10,
102  15, 21, 28, 36, 45, 55,
103  66, 78, 91, 105, 120, 136,
104  153, 171, 190, 210, 231, 253,
105  276, 300, 325, 351, 378, 406,
106  435, 465, 496, 528, 561, 595,
107  630, 666, 703, 741, 780, 820,
108  861, 903, 946, 990, 1035, 1081,
109  1128, 1176, 1225, 1275, 1326, 1378,
110  1431, 1485, 1540, 1596, 1653, 1711,
111  1770, 1830, 1891, 1953, 2016, 2080,
112  2145, 2211, 2278, 2346, 2415, 2485 },
113  { 0, 0, 0, 1, 4, 10,
114  20, 35, 56, 84, 120, 165,
115  220, 286, 364, 455, 560, 680,
116  816, 969, 1140, 1330, 1540, 1771,
117  2024, 2300, 2600, 2925, 3276, 3654,
118  4060, 4495, 4960, 5456, 5984, 6545,
119  7140, 7770, 8436, 9139, 9880, 10660,
120  11480, 12341, 13244, 14190, 15180, 16215,
121  17296, 18424, 19600, 20825, 22100, 23426,
122  24804, 26235, 27720, 29260, 30856, 32509,
123  34220, 35990, 37820, 39711, 41664, 43680,
124  45760, 47905, 50116, 52394, 54740, 57155 },
125  { 0, 0, 0, 0, 1, 5,
126  15, 35, 70, 126, 210, 330,
127  495, 715, 1001, 1365, 1820, 2380,
128  3060, 3876, 4845, 5985, 7315, 8855,
129  10626, 12650, 14950, 17550, 20475, 23751,
130  27405, 31465, 35960, 40920, 46376, 52360,
131  58905, 66045, 73815, 82251, 91390, 101270,
132  111930, 123410, 135751, 148995, 163185, 178365,
133  194580, 211876, 230300, 249900, 270725, 292825,
134  316251, 341055, 367290, 395010, 424270, 455126,
135  487635, 521855, 557845, 595665, 635376, 677040,
136  720720, 766480, 814385, 864501, 916895, 971635 },
137  { 0, 0, 0, 0, 0, 1,
138  6, 21, 56, 126, 252, 462,
139  792, 1287, 2002, 3003, 4368, 6188,
140  8568, 11628, 15504, 20349, 26334, 33649,
141  42504, 53130, 65780, 80730, 98280, 118755,
142  142506, 169911, 201376, 237336, 278256, 324632,
143  376992, 435897, 501942, 575757, 658008, 749398,
144  850668, 962598, 1086008, 1221759, 1370754, 1533939,
145  1712304, 1906884, 2118760, 2349060, 2598960, 2869685,
146  3162510, 3478761, 3819816, 4187106, 4582116, 5006386,
147  5461512, 5949147, 6471002, 7028847, 7624512, 8259888,
148  8936928, 9657648, 10424128, 11238513, 12103014, 13019909 },
149  { 0, 0, 0, 0, 0, 0,
150  1, 7, 28, 84, 210, 462,
151  924, 1716, 3003, 5005, 8008, 12376,
152  18564, 27132, 38760, 54264, 74613, 100947,
153  134596, 177100, 230230, 296010, 376740, 475020,
154  593775, 736281, 906192, 1107568, 1344904, 1623160,
155  1947792, 2324784, 2760681, 3262623, 3838380, 4496388,
156  5245786, 6096454, 7059052, 8145060, 9366819, 10737573,
157  12271512, 13983816, 15890700, 18009460, 20358520, 22957480,
158  25827165, 28989675, 32468436, 36288252, 40475358, 45057474,
159  50063860, 55525372, 61474519, 67945521, 74974368, 82598880,
160  90858768, 99795696, 109453344, 119877472, 131115985, 143218999 },
161  { 0, 0, 0, 0, 0, 0,
162  0, 1, 8, 36, 120, 330,
163  792, 1716, 3432, 6435, 11440, 19448,
164  31824, 50388, 77520, 116280, 170544, 245157,
165  346104, 480700, 657800, 888030, 1184040, 1560780,
166  2035800, 2629575, 3365856, 4272048, 5379616, 6724520,
167  8347680, 10295472, 12620256, 15380937, 18643560, 22481940,
168  26978328, 32224114, 38320568, 45379620, 53524680, 62891499,
169  73629072, 85900584, 99884400, 115775100, 133784560, 154143080,
170  177100560, 202927725, 231917400, 264385836, 300674088, 341149446,
171  386206920, 436270780, 491796152, 553270671, 621216192, 696190560,
172  778789440, 869648208, 969443904, 1078897248, 1198774720, 1329890705 },
173 };
174 
175 static const int16_t dss_sp_filter_cb[14][32] = {
176  { -32653, -32587, -32515, -32438, -32341, -32216, -32062, -31881,
177  -31665, -31398, -31080, -30724, -30299, -29813, -29248, -28572,
178  -27674, -26439, -24666, -22466, -19433, -16133, -12218, -7783,
179  -2834, 1819, 6544, 11260, 16050, 20220, 24774, 28120 },
180 
181  { -27503, -24509, -20644, -17496, -14187, -11277, -8420, -5595,
182  -3013, -624, 1711, 3880, 5844, 7774, 9739, 11592,
183  13364, 14903, 16426, 17900, 19250, 20586, 21803, 23006,
184  24142, 25249, 26275, 27300, 28359, 29249, 30118, 31183 },
185 
186  { -27827, -24208, -20943, -17781, -14843, -11848, -9066, -6297,
187  -3660, -910, 1918, 5025, 8223, 11649, 15086, 18423,
188  0, 0, 0, 0, 0, 0, 0, 0,
189  0, 0, 0, 0, 0, 0, 0, 0 },
190 
191  { -17128, -11975, -8270, -5123, -2296, 183, 2503, 4707,
192  6798, 8945, 11045, 13239, 15528, 18248, 21115, 24785,
193  0, 0, 0, 0, 0, 0, 0, 0,
194  0, 0, 0, 0, 0, 0, 0, 0 },
195 
196  { -21557, -17280, -14286, -11644, -9268, -7087, -4939, -2831,
197  -691, 1407, 3536, 5721, 8125, 10677, 13721, 17731,
198  0, 0, 0, 0, 0, 0, 0, 0,
199  0, 0, 0, 0, 0, 0, 0, 0 },
200 
201  { -15030, -10377, -7034, -4327, -1900, 364, 2458, 4450,
202  6422, 8374, 10374, 12486, 14714, 16997, 19626, 22954,
203  0, 0, 0, 0, 0, 0, 0, 0,
204  0, 0, 0, 0, 0, 0, 0, 0 },
205 
206  { -16155, -12362, -9698, -7460, -5258, -3359, -1547, 219,
207  1916, 3599, 5299, 6994, 8963, 11226, 13716, 16982,
208  0, 0, 0, 0, 0, 0, 0, 0,
209  0, 0, 0, 0, 0, 0, 0, 0 },
210 
211  { -14742, -9848, -6921, -4648, -2769, -1065, 499, 2083,
212  3633, 5219, 6857, 8580, 10410, 12672, 15561, 20101,
213  0, 0, 0, 0, 0, 0, 0, 0,
214  0, 0, 0, 0, 0, 0, 0, 0 },
215 
216  { -11099, -7014, -3855, -1025, 1680, 4544, 7807, 11932,
217  0, 0, 0, 0, 0, 0, 0, 0,
218  0, 0, 0, 0, 0, 0, 0, 0,
219  0, 0, 0, 0, 0, 0, 0, 0 },
220 
221  { -9060, -4570, -1381, 1419, 4034, 6728, 9865, 14149,
222  0, 0, 0, 0, 0, 0, 0, 0,
223  0, 0, 0, 0, 0, 0, 0, 0,
224  0, 0, 0, 0, 0, 0, 0, 0 },
225 
226  { -12450, -7985, -4596, -1734, 961, 3629, 6865, 11142,
227  0, 0, 0, 0, 0, 0, 0, 0,
228  0, 0, 0, 0, 0, 0, 0, 0,
229  0, 0, 0, 0, 0, 0, 0, 0 },
230 
231  { -11831, -7404, -4010, -1096, 1606, 4291, 7386, 11482,
232  0, 0, 0, 0, 0, 0, 0, 0,
233  0, 0, 0, 0, 0, 0, 0, 0,
234  0, 0, 0, 0, 0, 0, 0, 0 },
235 
236  { -13404, -9250, -5995, -3312, -890, 1594, 4464, 8198,
237  0, 0, 0, 0, 0, 0, 0, 0,
238  0, 0, 0, 0, 0, 0, 0, 0,
239  0, 0, 0, 0, 0, 0, 0, 0 },
240 
241  { -11239, -7220, -4040, -1406, 971, 3321, 6006, 9697,
242  0, 0, 0, 0, 0, 0, 0, 0,
243  0, 0, 0, 0, 0, 0, 0, 0,
244  0, 0, 0, 0, 0, 0, 0, 0 },
245 };
246 
247 static const uint16_t dss_sp_fixed_cb_gain[64] = {
248  0, 4, 8, 13, 17, 22, 26, 31,
249  35, 40, 44, 48, 53, 58, 63, 69,
250  76, 83, 91, 99, 109, 119, 130, 142,
251  155, 170, 185, 203, 222, 242, 265, 290,
252  317, 346, 378, 414, 452, 494, 540, 591,
253  646, 706, 771, 843, 922, 1007, 1101, 1204,
254  1316, 1438, 1572, 1719, 1879, 2053, 2244, 2453,
255  2682, 2931, 3204, 3502, 3828, 4184, 4574, 5000,
256 };
257 
258 static const int16_t dss_sp_pulse_val[8] = {
259  -31182, -22273, -13364, -4455, 4455, 13364, 22273, 31182
260 };
261 
262 static const uint16_t binary_decreasing_array[] = {
263  32767, 16384, 8192, 4096, 2048, 1024, 512, 256,
264  128, 64, 32, 16, 8, 4, 2,
265 };
266 
267 static const uint16_t dss_sp_unc_decreasing_array[] = {
268  32767, 26214, 20972, 16777, 13422, 10737, 8590, 6872,
269  5498, 4398, 3518, 2815, 2252, 1801, 1441,
270 };
271 
272 static const uint16_t dss_sp_adaptive_gain[] = {
273  102, 231, 360, 488, 617, 746, 875, 1004,
274  1133, 1261, 1390, 1519, 1648, 1777, 1905, 2034,
275  2163, 2292, 2421, 2550, 2678, 2807, 2936, 3065,
276  3194, 3323, 3451, 3580, 3709, 3838, 3967, 4096,
277 };
278 
279 static const int32_t dss_sp_sinc[67] = {
280  262, 293, 323, 348, 356, 336, 269, 139,
281  -67, -358, -733, -1178, -1668, -2162, -2607, -2940,
282  -3090, -2986, -2562, -1760, -541, 1110, 3187, 5651,
283  8435, 11446, 14568, 17670, 20611, 23251, 25460, 27125,
284  28160, 28512, 28160,
285  27125, 25460, 23251, 20611, 17670, 14568, 11446, 8435,
286  5651, 3187, 1110, -541, -1760, -2562, -2986, -3090,
287  -2940, -2607, -2162, -1668, -1178, -733, -358, -67,
288  139, 269, 336, 356, 348, 323, 293, 262,
289 };
290 
291 static av_cold int dss_sp_decode_init(AVCodecContext *avctx)
292 {
293  DssSpContext *p = avctx->priv_data;
294  avctx->channel_layout = AV_CH_LAYOUT_MONO;
295  avctx->sample_fmt = AV_SAMPLE_FMT_S16;
296  avctx->channels = 1;
297  avctx->sample_rate = 11025;
298 
299  memset(p->history, 0, sizeof(p->history));
300  p->pulse_dec_mode = 1;
301  p->avctx = avctx;
302 
303  return 0;
304 }
305 
306 static void dss_sp_unpack_coeffs(DssSpContext *p, const uint8_t *src)
307 {
308  GetBitContext gb;
309  DssSpFrame *fparam = &p->fparam;
310  int i;
311  int subframe_idx;
312  uint32_t combined_pitch;
313  uint32_t tmp;
314  uint32_t pitch_lag;
315 
316  for (i = 0; i < DSS_SP_FRAME_SIZE; i += 2) {
317  p->bits[i] = src[i + 1];
318  p->bits[i + 1] = src[i];
319  }
320 
321  init_get_bits(&gb, p->bits, DSS_SP_FRAME_SIZE * 8);
322 
323  for (i = 0; i < 2; i++)
324  fparam->filter_idx[i] = get_bits(&gb, 5);
325  for (; i < 8; i++)
326  fparam->filter_idx[i] = get_bits(&gb, 4);
327  for (; i < 14; i++)
328  fparam->filter_idx[i] = get_bits(&gb, 3);
329 
330  for (subframe_idx = 0; subframe_idx < 4; subframe_idx++) {
331  fparam->sf_adaptive_gain[subframe_idx] = get_bits(&gb, 5);
332 
333  fparam->sf[subframe_idx].combined_pulse_pos = get_bits_long(&gb, 31);
334 
335  fparam->sf[subframe_idx].gain = get_bits(&gb, 6);
336 
337  for (i = 0; i < 7; i++)
338  fparam->sf[subframe_idx].pulse_val[i] = get_bits(&gb, 3);
339  }
340 
341  for (subframe_idx = 0; subframe_idx < 4; subframe_idx++) {
342  unsigned int C72_binomials[PULSE_MAX] = {
343  72, 2556, 59640, 1028790, 13991544, 156238908, 1473109704,
344  3379081753
345  };
346  unsigned int combined_pulse_pos =
347  fparam->sf[subframe_idx].combined_pulse_pos;
348  int index = 6;
349 
350  if (combined_pulse_pos < C72_binomials[PULSE_MAX - 1]) {
351  if (p->pulse_dec_mode) {
352  int pulse, pulse_idx;
353  pulse = PULSE_MAX - 1;
354  pulse_idx = 71;
355  combined_pulse_pos =
356  fparam->sf[subframe_idx].combined_pulse_pos;
357 
358  /* this part seems to be close to g723.1 gen_fcb_excitation()
359  * RATE_6300 */
360 
361  /* TODO: what is 7? size of subframe? */
362  for (i = 0; i < 7; i++) {
363  for (;
364  combined_pulse_pos <
365  dss_sp_combinatorial_table[pulse][pulse_idx];
366  --pulse_idx)
367  ;
368  combined_pulse_pos -=
369  dss_sp_combinatorial_table[pulse][pulse_idx];
370  pulse--;
371  fparam->sf[subframe_idx].pulse_pos[i] = pulse_idx;
372  }
373  }
374  } else {
375  p->pulse_dec_mode = 0;
376 
377  /* why do we need this? */
378  fparam->sf[subframe_idx].pulse_pos[6] = 0;
379 
380  for (i = 71; i >= 0; i--) {
381  if (C72_binomials[index] <= combined_pulse_pos) {
382  combined_pulse_pos -= C72_binomials[index];
383 
384  fparam->sf[subframe_idx].pulse_pos[6 - index] = i;
385 
386  if (!index)
387  break;
388  --index;
389  }
390  --C72_binomials[0];
391  if (index) {
392  int a;
393  for (a = 0; a < index; a++)
394  C72_binomials[a + 1] -= C72_binomials[a];
395  }
396  }
397  }
398  }
399 
400  combined_pitch = get_bits(&gb, 24);
401 
402  fparam->pitch_lag[0] = (combined_pitch % 151) + 36;
403 
404  combined_pitch /= 151;
405 
406  for (i = 1; i < SUBFRAMES - 1; i++) {
407  fparam->pitch_lag[i] = combined_pitch % 48;
408  combined_pitch /= 48;
409  }
410  if (combined_pitch > 47) {
411  av_log (p->avctx, AV_LOG_WARNING, "combined_pitch was too large\n");
412  combined_pitch = 0;
413  }
414  fparam->pitch_lag[i] = combined_pitch;
415 
416  pitch_lag = fparam->pitch_lag[0];
417  for (i = 1; i < SUBFRAMES; i++) {
418  if (pitch_lag > 162) {
419  fparam->pitch_lag[i] += 162 - 23;
420  } else {
421  tmp = pitch_lag - 23;
422  if (tmp < 36)
423  tmp = 36;
424  fparam->pitch_lag[i] += tmp;
425  }
426  pitch_lag = fparam->pitch_lag[i];
427  }
428 }
429 
430 static void dss_sp_unpack_filter(DssSpContext *p)
431 {
432  int i;
433 
434  for (i = 0; i < 14; i++)
435  p->lpc_filter[i] = dss_sp_filter_cb[i][p->fparam.filter_idx[i]];
436 }
437 
438 static void dss_sp_convert_coeffs(int32_t *lpc_filter, int32_t *coeffs)
439 {
440  int a, a_plus, i;
441 
442  coeffs[0] = 0x2000;
443  for (a = 0; a < 14; a++) {
444  a_plus = a + 1;
445  coeffs[a_plus] = lpc_filter[a] >> 2;
446  if (a_plus / 2 >= 1) {
447  for (i = 1; i <= a_plus / 2; i++) {
448  int coeff_1, coeff_2, tmp;
449 
450  coeff_1 = coeffs[i];
451  coeff_2 = coeffs[a_plus - i];
452 
453  tmp = DSS_SP_FORMULA(coeff_1, lpc_filter[a], coeff_2);
454  coeffs[i] = av_clip_int16(tmp);
455 
456  tmp = DSS_SP_FORMULA(coeff_2, lpc_filter[a], coeff_1);
457  coeffs[a_plus - i] = av_clip_int16(tmp);
458  }
459  }
460  }
461 }
462 
463 static void dss_sp_add_pulses(int32_t *vector_buf,
464  const struct DssSpSubframe *sf)
465 {
466  int i;
467 
468  for (i = 0; i < 7; i++)
469  vector_buf[sf->pulse_pos[i]] += (dss_sp_fixed_cb_gain[sf->gain] *
470  dss_sp_pulse_val[sf->pulse_val[i]] +
471  0x4000) >> 15;
472 }
473 
474 static void dss_sp_gen_exc(int32_t *vector, int32_t *prev_exc,
475  int pitch_lag, int gain)
476 {
477  int i;
478 
479  /* do we actually need this check? we can use just [a3 - i % a3]
480  * for both cases */
481  if (pitch_lag < 72)
482  for (i = 0; i < 72; i++)
483  vector[i] = prev_exc[pitch_lag - i % pitch_lag];
484  else
485  for (i = 0; i < 72; i++)
486  vector[i] = prev_exc[pitch_lag - i];
487 
488  for (i = 0; i < 72; i++) {
489  int tmp = gain * vector[i] >> 11;
490  vector[i] = av_clip_int16(tmp);
491  }
492 }
493 
494 static void dss_sp_scale_vector(int32_t *vec, int bits, int size)
495 {
496  int i;
497 
498  if (bits < 0)
499  for (i = 0; i < size; i++)
500  vec[i] = vec[i] >> -bits;
501  else
502  for (i = 0; i < size; i++)
503  vec[i] = vec[i] * (1 << bits);
504 }
505 
506 static void dss_sp_update_buf(int32_t *hist, int32_t *vector)
507 {
508  int i;
509 
510  for (i = 114; i > 0; i--)
511  vector[i + 72] = vector[i];
512 
513  for (i = 0; i < 72; i++)
514  vector[72 - i] = hist[i];
515 }
516 
517 static void dss_sp_shift_sq_sub(const int32_t *filter_buf,
518  int32_t *error_buf, int32_t *dst)
519 {
520  int a;
521 
522  for (a = 0; a < 72; a++) {
523  int i, tmp;
524 
525  tmp = dst[a] * filter_buf[0];
526 
527  for (i = 14; i > 0; i--)
528  tmp -= error_buf[i] * (unsigned)filter_buf[i];
529 
530  for (i = 14; i > 0; i--)
531  error_buf[i] = error_buf[i - 1];
532 
533  tmp = (int)(tmp + 4096U) >> 13;
534 
535  error_buf[1] = tmp;
536 
537  dst[a] = av_clip_int16(tmp);
538  }
539 }
540 
541 static void dss_sp_shift_sq_add(const int32_t *filter_buf, int32_t *audio_buf,
542  int32_t *dst)
543 {
544  int a;
545 
546  for (a = 0; a < 72; a++) {
547  int i, tmp = 0;
548 
549  audio_buf[0] = dst[a];
550 
551  for (i = 14; i >= 0; i--)
552  tmp += audio_buf[i] * filter_buf[i];
553 
554  for (i = 14; i > 0; i--)
555  audio_buf[i] = audio_buf[i - 1];
556 
557  tmp = (tmp + 4096) >> 13;
558 
559  dst[a] = av_clip_int16(tmp);
560  }
561 }
562 
563 static void dss_sp_vec_mult(const int32_t *src, int32_t *dst,
564  const int16_t *mult)
565 {
566  int i;
567 
568  dst[0] = src[0];
569 
570  for (i = 1; i < 15; i++)
571  dst[i] = (src[i] * mult[i] + 0x4000) >> 15;
572 }
573 
574 static int dss_sp_get_normalize_bits(int32_t *vector_buf, int16_t size)
575 {
576  unsigned int val;
577  int max_val;
578  int i;
579 
580  val = 1;
581  for (i = 0; i < size; i++)
582  val |= FFABS(vector_buf[i]);
583 
584  for (max_val = 0; val <= 0x4000; ++max_val)
585  val *= 2;
586  return max_val;
587 }
588 
589 static int dss_sp_vector_sum(DssSpContext *p, int size)
590 {
591  int i, sum = 0;
592  for (i = 0; i < size; i++)
593  sum += FFABS(p->vector_buf[i]);
594  return sum;
595 }
596 
597 static void dss_sp_sf_synthesis(DssSpContext *p, int32_t lpc_filter,
598  int32_t *dst, int size)
599 {
600  int32_t tmp_buf[15];
601  int32_t noise[72];
602  int bias, vsum_2 = 0, vsum_1 = 0, v36, normalize_bits;
603  int i, tmp;
604 
605  if (size > 0) {
606  vsum_1 = dss_sp_vector_sum(p, size);
607 
608  if (vsum_1 > 0xFFFFF)
609  vsum_1 = 0xFFFFF;
610  }
611 
612  normalize_bits = dss_sp_get_normalize_bits(p->vector_buf, size);
613 
614  dss_sp_scale_vector(p->vector_buf, normalize_bits - 3, size);
615  dss_sp_scale_vector(p->audio_buf, normalize_bits, 15);
616  dss_sp_scale_vector(p->err_buf1, normalize_bits, 15);
617 
618  v36 = p->err_buf1[1];
619 
620  dss_sp_vec_mult(p->filter, tmp_buf, binary_decreasing_array);
621  dss_sp_shift_sq_add(tmp_buf, p->audio_buf, p->vector_buf);
622 
623  dss_sp_vec_mult(p->filter, tmp_buf, dss_sp_unc_decreasing_array);
624  dss_sp_shift_sq_sub(tmp_buf, p->err_buf1, p->vector_buf);
625 
626  /* lpc_filter can be negative */
627  lpc_filter = lpc_filter >> 1;
628  if (lpc_filter >= 0)
629  lpc_filter = 0;
630 
631  if (size > 1) {
632  for (i = size - 1; i > 0; i--) {
633  tmp = DSS_SP_FORMULA(p->vector_buf[i], lpc_filter,
634  p->vector_buf[i - 1]);
635  p->vector_buf[i] = av_clip_int16(tmp);
636  }
637  }
638 
639  tmp = DSS_SP_FORMULA(p->vector_buf[0], lpc_filter, v36);
640  p->vector_buf[0] = av_clip_int16(tmp);
641 
642  dss_sp_scale_vector(p->vector_buf, -normalize_bits, size);
643  dss_sp_scale_vector(p->audio_buf, -normalize_bits, 15);
644  dss_sp_scale_vector(p->err_buf1, -normalize_bits, 15);
645 
646  if (size > 0)
647  vsum_2 = dss_sp_vector_sum(p, size);
648 
649  if (vsum_2 >= 0x40)
650  tmp = (vsum_1 << 11) / vsum_2;
651  else
652  tmp = 1;
653 
654  bias = 409 * tmp >> 15 << 15;
655  tmp = (bias + 32358 * p->noise_state) >> 15;
656  noise[0] = av_clip_int16(tmp);
657 
658  for (i = 1; i < size; i++) {
659  tmp = (bias + 32358 * noise[i - 1]) >> 15;
660  noise[i] = av_clip_int16(tmp);
661  }
662 
663  p->noise_state = noise[size - 1];
664  for (i = 0; i < size; i++) {
665  tmp = (p->vector_buf[i] * noise[i]) >> 11;
666  dst[i] = av_clip_int16(tmp);
667  }
668 }
669 
670 static void dss_sp_update_state(DssSpContext *p, int32_t *dst)
671 {
672  int i, offset = 6, counter = 0, a = 0;
673 
674  for (i = 0; i < 6; i++)
675  p->excitation[i] = p->excitation[288 + i];
676 
677  for (i = 0; i < 72 * SUBFRAMES; i++)
678  p->excitation[6 + i] = dst[i];
679 
680  do {
681  int tmp = 0;
682 
683  for (i = 0; i < 6; i++)
684  tmp += p->excitation[offset--] * dss_sp_sinc[a + i * 11];
685 
686  offset += 7;
687 
688  tmp >>= 15;
689  dst[counter] = av_clip_int16(tmp);
690 
691  counter++;
692 
693  a = (a + 1) % 11;
694  if (!a)
695  offset++;
696  } while (offset < FF_ARRAY_ELEMS(p->excitation));
697 }
698 
699 static void dss_sp_32to16bit(int16_t *dst, int32_t *src, int size)
700 {
701  int i;
702 
703  for (i = 0; i < size; i++)
704  dst[i] = av_clip_int16(src[i]);
705 }
706 
707 static int dss_sp_decode_one_frame(DssSpContext *p,
708  int16_t *abuf_dst, const uint8_t *abuf_src)
709 {
710  int i, j;
711 
712  dss_sp_unpack_coeffs(p, abuf_src);
713 
714  dss_sp_unpack_filter(p);
715 
716  dss_sp_convert_coeffs(p->lpc_filter, p->filter);
717 
718  for (j = 0; j < SUBFRAMES; j++) {
719  dss_sp_gen_exc(p->vector_buf, p->history,
720  p->fparam.pitch_lag[j],
721  dss_sp_adaptive_gain[p->fparam.sf_adaptive_gain[j]]);
722 
723  dss_sp_add_pulses(p->vector_buf, &p->fparam.sf[j]);
724 
725  dss_sp_update_buf(p->vector_buf, p->history);
726 
727  for (i = 0; i < 72; i++)
728  p->vector_buf[i] = p->history[72 - i];
729 
730  dss_sp_shift_sq_sub(p->filter,
731  p->err_buf2, p->vector_buf);
732 
733  dss_sp_sf_synthesis(p, p->lpc_filter[0],
734  &p->working_buffer[j][0], 72);
735  }
736 
737  dss_sp_update_state(p, &p->working_buffer[0][0]);
738 
739  dss_sp_32to16bit(abuf_dst,
740  &p->working_buffer[0][0], 264);
741  return 0;
742 }
743 
744 static int dss_sp_decode_frame(AVCodecContext *avctx, void *data,
745  int *got_frame_ptr, AVPacket *avpkt)
746 {
747  DssSpContext *p = avctx->priv_data;
748  AVFrame *frame = data;
749  const uint8_t *buf = avpkt->data;
750  int buf_size = avpkt->size;
751 
752  int16_t *out;
753  int ret;
754 
755  if (buf_size < DSS_SP_FRAME_SIZE) {
756  if (buf_size)
757  av_log(avctx, AV_LOG_WARNING,
758  "Expected %d bytes, got %d - skipping packet.\n",
759  DSS_SP_FRAME_SIZE, buf_size);
760  *got_frame_ptr = 0;
761  return AVERROR_INVALIDDATA;
762  }
763 
764  frame->nb_samples = DSS_SP_SAMPLE_COUNT;
765  if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
766  return ret;
767 
768  out = (int16_t *)frame->data[0];
769 
770  dss_sp_decode_one_frame(p, out, buf);
771 
772  *got_frame_ptr = 1;
773 
774  return DSS_SP_FRAME_SIZE;
775 }
776 
777 AVCodec ff_dss_sp_decoder = {
778  .name = "dss_sp",
779  .long_name = NULL_IF_CONFIG_SMALL("Digital Speech Standard - Standard Play mode (DSS SP)"),
780  .type = AVMEDIA_TYPE_AUDIO,
781  .id = AV_CODEC_ID_DSS_SP,
782  .priv_data_size = sizeof(DssSpContext),
783  .init = dss_sp_decode_init,
784  .decode = dss_sp_decode_frame,
785  .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_CHANNEL_CONF,
786 };
AVCodec
AVCodec.
Definition: codec.h:197
AV_LOG_WARNING
#define AV_LOG_WARNING
Something somehow does not look correct.
Definition: log.h:200
init
static av_cold int init(AVCodecContext *avctx)
Definition: avrndec.c:31
dss_sp_adaptive_gain
static const uint16_t dss_sp_adaptive_gain[]
Definition: dss_sp.c:272
opt.h
DssSpContext::vector_buf
int32_t vector_buf[72]
Definition: dss_sp.c:63
DssSpSubframe::gain
int16_t gain
Definition: dss_sp.c:40
AVCodecContext::channel_layout
uint64_t channel_layout
Audio channel layout.
Definition: avcodec.h:1247
mem_internal.h
out
FILE * out
Definition: movenc.c:54
AVCodecContext::sample_rate
int sample_rate
samples per second
Definition: avcodec.h:1196
dss_sp_32to16bit
static void dss_sp_32to16bit(int16_t *dst, int32_t *src, int size)
Definition: dss_sp.c:699
AV_CH_LAYOUT_MONO
#define AV_CH_LAYOUT_MONO
Definition: channel_layout.h:90
get_bits_long
static unsigned int get_bits_long(GetBitContext *s, int n)
Read 0-32 bits.
Definition: get_bits.h:546
DssSpContext::avctx
AVCodecContext * avctx
Definition: dss_sp.c:54
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:324
tmp
static uint8_t tmp[11]
Definition: aes_ctr.c:27
index
fg index
Definition: ffmpeg_filter.c:170
internal.h
AVPacket::data
uint8_t * data
Definition: packet.h:369
data
const char data[16]
Definition: mxf.c:142
DssSpContext::pulse_dec_mode
int pulse_dec_mode
Definition: dss_sp.c:67
init_get_bits
static int init_get_bits(GetBitContext *s, const uint8_t *buffer, int bit_size)
Initialize GetBitContext.
Definition: get_bits.h:659
DssSpContext::lpc_filter
int32_t lpc_filter[14]
Definition: dss_sp.c:61
DssSpContext::noise_state
int noise_state
Definition: dss_sp.c:64
get_bits
static unsigned int get_bits(GetBitContext *s, int n)
Read 1-25 bits.
Definition: get_bits.h:379
dss_sp_convert_coeffs
static void dss_sp_convert_coeffs(int32_t *lpc_filter, int32_t *coeffs)
Definition: dss_sp.c:438
DssSpContext::audio_buf
int32_t audio_buf[15]
Definition: dss_sp.c:59
U
#define U(x)
Definition: vp56_arith.h:37
DssSpContext::filter
int32_t filter[15]
Definition: dss_sp.c:62
GetBitContext
Definition: get_bits.h:61
val
static double val(void *priv, double ch)
Definition: aeval.c:76
dss_sp_combinatorial_table
static const uint32_t dss_sp_combinatorial_table[PULSE_MAX][72]
Definition: dss_sp.c:76
dss_sp_shift_sq_add
static void dss_sp_shift_sq_add(const int32_t *filter_buf, int32_t *audio_buf, int32_t *dst)
Definition: dss_sp.c:541
mult
static int16_t mult(Float11 *f1, Float11 *f2)
Definition: g726.c:55
DssSpSubframe::pulse_pos
int16_t pulse_pos[7]
Definition: dss_sp.c:42
FF_ARRAY_ELEMS
#define FF_ARRAY_ELEMS(a)
Definition: sinewin_tablegen.c:29
av_cold
#define av_cold
Definition: attributes.h:90
decode
static void decode(AVCodecContext *dec_ctx, AVPacket *pkt, AVFrame *frame, FILE *outfile)
Definition: decode_audio.c:71
dss_sp_update_buf
static void dss_sp_update_buf(int32_t *hist, int32_t *vector)
Definition: dss_sp.c:506
AVMEDIA_TYPE_AUDIO
@ AVMEDIA_TYPE_AUDIO
Definition: avutil.h:202
DssSpSubframe::combined_pulse_pos
int32_t combined_pulse_pos
Definition: dss_sp.c:41
dss_sp_unpack_filter
static void dss_sp_unpack_filter(DssSpContext *p)
Definition: dss_sp.c:430
bits
uint8_t bits
Definition: vp3data.h:141
dss_sp_update_state
static void dss_sp_update_state(DssSpContext *p, int32_t *dst)
Definition: dss_sp.c:670
DSS_SP_FRAME_SIZE
#define DSS_SP_FRAME_SIZE
Definition: dss_sp.c:35
get_bits.h
DSS_SP_SAMPLE_COUNT
#define DSS_SP_SAMPLE_COUNT
Definition: dss_sp.c:36
dss_sp_decode_init
static av_cold int dss_sp_decode_init(AVCodecContext *avctx)
Definition: dss_sp.c:291
int32_t
int32_t
Definition: audio_convert.c:194
FFABS
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:72
DssSpContext::fparam
DssSpFrame fparam
Definition: dss_sp.c:57
av_clip_int16
#define av_clip_int16
Definition: common.h:137
DssSpFrame::sf
struct DssSpSubframe sf[SUBFRAMES]
Definition: dss_sp.c:50
dss_sp_add_pulses
static void dss_sp_add_pulses(int32_t *vector_buf, const struct DssSpSubframe *sf)
Definition: dss_sp.c:463
src
#define src
Definition: vp8dsp.c:255
SUBFRAMES
#define SUBFRAMES
Definition: dss_sp.c:32
dss_sp_get_normalize_bits
static int dss_sp_get_normalize_bits(int32_t *vector_buf, int16_t size)
Definition: dss_sp.c:574
dss_sp_filter_cb
static const int16_t dss_sp_filter_cb[14][32]
Definition: dss_sp.c:175
dss_sp_vec_mult
static void dss_sp_vec_mult(const int32_t *src, int32_t *dst, const int16_t *mult)
Definition: dss_sp.c:563
for
for(j=16;j >0;--j)
Definition: h264pred_template.c:469
AV_CODEC_CAP_CHANNEL_CONF
#define AV_CODEC_CAP_CHANNEL_CONF
Codec should fill in channel configuration and samplerate instead of container.
Definition: codec.h:104
dss_sp_vector_sum
static int dss_sp_vector_sum(DssSpContext *p, int size)
Definition: dss_sp.c:589
DssSpContext
Definition: dss_sp.c:53
DssSpContext::history
int32_t history[187]
Definition: dss_sp.c:56
ff_get_buffer
int ff_get_buffer(AVCodecContext *avctx, AVFrame *frame, int flags)
Get a buffer for a frame.
Definition: decode.c:1893
AV_CODEC_CAP_DR1
#define AV_CODEC_CAP_DR1
Codec uses get_buffer() or get_encode_buffer() for allocating buffers and supports custom allocators.
Definition: codec.h:52
AVPacket::size
int size
Definition: packet.h:370
NULL_IF_CONFIG_SMALL
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:117
DssSpFrame
Definition: dss_sp.c:46
dss_sp_sinc
static const int32_t dss_sp_sinc[67]
Definition: dss_sp.c:279
AVCodecContext::sample_fmt
enum AVSampleFormat sample_fmt
audio sample format
Definition: avcodec.h:1204
size
int size
Definition: twinvq_data.h:10344
a
The reader does not expect b to be semantically here and if the code is changed by maybe adding a a division or other the signedness will almost certainly be mistaken To avoid this confusion a new type was SUINT is the C unsigned type but it holds a signed int to use the same example SUINT a
Definition: undefined.txt:41
offset
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf offset
Definition: writing_filters.txt:86
dss_sp_scale_vector
static void dss_sp_scale_vector(int32_t *vec, int bits, int size)
Definition: dss_sp.c:494
dss_sp_decode_frame
static int dss_sp_decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr, AVPacket *avpkt)
Definition: dss_sp.c:744
DssSpSubframe
Definition: dss_sp.c:39
AVCodecContext::channels
int channels
number of audio channels
Definition: avcodec.h:1197
DSS_SP_FORMULA
#define DSS_SP_FORMULA(a, b, c)
Definition: dss_sp.c:37
DECLARE_ALIGNED
#define DECLARE_ALIGNED(n, t, v)
Definition: mem.h:117
dss_sp_unpack_coeffs
static void dss_sp_unpack_coeffs(DssSpContext *p, const uint8_t *src)
Definition: dss_sp.c:306
DssSpSubframe::pulse_val
int16_t pulse_val[7]
Definition: dss_sp.c:43
i
int i
Definition: input.c:407
common.h
DssSpFrame::sf_adaptive_gain
int16_t sf_adaptive_gain[SUBFRAMES]
Definition: dss_sp.c:48
uint8_t
uint8_t
Definition: audio_convert.c:194
DssSpContext::excitation
int32_t excitation[288+6]
Definition: dss_sp.c:55
AV_CODEC_ID_DSS_SP
@ AV_CODEC_ID_DSS_SP
Definition: codec_id.h:490
AV_SAMPLE_FMT_S16
@ AV_SAMPLE_FMT_S16
signed 16 bits
Definition: samplefmt.h:61
AVCodec::name
const char * name
Name of the codec implementation.
Definition: codec.h:204
avcodec.h
DssSpContext::bits
uint8_t bits[DSS_SP_FRAME_SIZE+AV_INPUT_BUFFER_PADDING_SIZE]
Definition: dss_sp.c:70
ret
ret
Definition: filter_design.txt:187
frame
these buffered frames must be flushed immediately if a new input produces new the filter must not call request_frame to get more It must just process the frame or queue it The task of requesting more frames is left to the filter s request_frame method or the application If a filter has several the filter must be ready for frames arriving randomly on any input any filter with several inputs will most likely require some kind of queuing mechanism It is perfectly acceptable to have a limited queue and to drop frames when the inputs are too unbalanced request_frame For filters that do not use the this method is called when a frame is wanted on an output For a it should directly call filter_frame on the corresponding output For a if there are queued frames already one of these frames should be pushed If the filter should request a frame on one of its repeatedly until at least one frame has been pushed Return or at least make progress towards producing a frame
Definition: filter_design.txt:264
dss_sp_shift_sq_sub
static void dss_sp_shift_sq_sub(const int32_t *filter_buf, int32_t *error_buf, int32_t *dst)
Definition: dss_sp.c:517
dss_sp_decode_one_frame
static int dss_sp_decode_one_frame(DssSpContext *p, int16_t *abuf_dst, const uint8_t *abuf_src)
Definition: dss_sp.c:707
AV_INPUT_BUFFER_PADDING_SIZE
#define AV_INPUT_BUFFER_PADDING_SIZE
Definition: avcodec.h:215
dss_sp_unc_decreasing_array
static const uint16_t dss_sp_unc_decreasing_array[]
Definition: dss_sp.c:267
AVCodecContext
main external API structure.
Definition: avcodec.h:536
channel_layout.h
noise
static int noise(AVBSFContext *ctx, AVPacket *pkt)
Definition: noise_bsf.c:36
PULSE_MAX
#define PULSE_MAX
Definition: dss_sp.c:33
mem.h
dss_sp_sf_synthesis
static void dss_sp_sf_synthesis(DssSpContext *p, int32_t lpc_filter, int32_t *dst, int size)
Definition: dss_sp.c:597
DssSpFrame::pitch_lag
int16_t pitch_lag[SUBFRAMES]
Definition: dss_sp.c:49
AVPacket
This structure stores compressed data.
Definition: packet.h:346
AVCodecContext::priv_data
void * priv_data
Definition: avcodec.h:563
DssSpContext::working_buffer
int32_t working_buffer[SUBFRAMES][72]
Definition: dss_sp.c:58
DssSpContext::err_buf2
int32_t err_buf2[15]
Definition: dss_sp.c:65
dss_sp_pulse_val
static const int16_t dss_sp_pulse_val[8]
Definition: dss_sp.c:258
ff_dss_sp_decoder
AVCodec ff_dss_sp_decoder
Definition: dss_sp.c:777
binary_decreasing_array
static const uint16_t binary_decreasing_array[]
Definition: dss_sp.c:262
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:28
AVERROR_INVALIDDATA
#define AVERROR_INVALIDDATA
Invalid data found when processing input.
Definition: error.h:59
int
int
Definition: ffmpeg_filter.c:158
DssSpContext::err_buf1
int32_t err_buf1[15]
Definition: dss_sp.c:60
dss_sp_gen_exc
static void dss_sp_gen_exc(int32_t *vector, int32_t *prev_exc, int pitch_lag, int gain)
Definition: dss_sp.c:474
DssSpFrame::filter_idx
int16_t filter_idx[14]
Definition: dss_sp.c:47
dss_sp_fixed_cb_gain
static const uint16_t dss_sp_fixed_cb_gain[64]
Definition: dss_sp.c:247
Generated on Mon Apr 19 2021 19:21:13 for FFmpeg by   doxygen 1.8.17