g726.c
Go to the documentation of this file.
1 /*
2  * G.726 ADPCM audio codec
3  * Copyright (c) 2004 Roman Shaposhnik
4  *
5  * This is a very straightforward rendition of the G.726
6  * Section 4 "Computational Details".
7  *
8  * This file is part of Libav.
9  *
10  * Libav is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU Lesser General Public
12  * License as published by the Free Software Foundation; either
13  * version 2.1 of the License, or (at your option) any later version.
14  *
15  * Libav is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18  * Lesser General Public License for more details.
19  *
20  * You should have received a copy of the GNU Lesser General Public
21  * License along with Libav; if not, write to the Free Software
22  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23  */
24 #include <limits.h>
25 #include "libavutil/avassert.h"
26 #include "libavutil/opt.h"
27 #include "avcodec.h"
28 #include "internal.h"
29 #include "get_bits.h"
30 #include "put_bits.h"
31 
38 typedef struct Float11 {
39  uint8_t sign;
40  uint8_t exp;
41  uint8_t mant;
42 } Float11;
43 
44 static inline Float11* i2f(int i, Float11* f)
45 {
46  f->sign = (i < 0);
47  if (f->sign)
48  i = -i;
49  f->exp = av_log2_16bit(i) + !!i;
50  f->mant = i? (i<<6) >> f->exp : 1<<5;
51  return f;
52 }
53 
54 static inline int16_t mult(Float11* f1, Float11* f2)
55 {
56  int res, exp;
57 
58  exp = f1->exp + f2->exp;
59  res = (((f1->mant * f2->mant) + 0x30) >> 4);
60  res = exp > 19 ? res << (exp - 19) : res >> (19 - exp);
61  return (f1->sign ^ f2->sign) ? -res : res;
62 }
63 
64 static inline int sgn(int value)
65 {
66  return (value < 0) ? -1 : 1;
67 }
68 
69 typedef struct G726Tables {
70  const int* quant;
71  const int16_t* iquant;
72  const int16_t* W;
73  const uint8_t* F;
74 } G726Tables;
75 
76 typedef struct G726Context {
77  AVClass *class;
81  Float11 sr[2];
82  Float11 dq[6];
83  int a[2];
84  int b[6];
85  int pk[2];
87  int ap;
88  int yu;
89  int yl;
90  int dms;
91  int dml;
92  int td;
94  int se;
95  int sez;
96  int y;
97  int code_size;
98 } G726Context;
99 
100 static const int quant_tbl16[] =
101  { 260, INT_MAX };
102 static const int16_t iquant_tbl16[] =
103  { 116, 365, 365, 116 };
104 static const int16_t W_tbl16[] =
105  { -22, 439, 439, -22 };
106 static const uint8_t F_tbl16[] =
107  { 0, 7, 7, 0 };
108 
109 static const int quant_tbl24[] =
110  { 7, 217, 330, INT_MAX };
111 static const int16_t iquant_tbl24[] =
112  { INT16_MIN, 135, 273, 373, 373, 273, 135, INT16_MIN };
113 static const int16_t W_tbl24[] =
114  { -4, 30, 137, 582, 582, 137, 30, -4 };
115 static const uint8_t F_tbl24[] =
116  { 0, 1, 2, 7, 7, 2, 1, 0 };
117 
118 static const int quant_tbl32[] =
119  { -125, 79, 177, 245, 299, 348, 399, INT_MAX };
120 static const int16_t iquant_tbl32[] =
121  { INT16_MIN, 4, 135, 213, 273, 323, 373, 425,
122  425, 373, 323, 273, 213, 135, 4, INT16_MIN };
123 static const int16_t W_tbl32[] =
124  { -12, 18, 41, 64, 112, 198, 355, 1122,
125  1122, 355, 198, 112, 64, 41, 18, -12};
126 static const uint8_t F_tbl32[] =
127  { 0, 0, 0, 1, 1, 1, 3, 7, 7, 3, 1, 1, 1, 0, 0, 0 };
128 
129 static const int quant_tbl40[] =
130  { -122, -16, 67, 138, 197, 249, 297, 338,
131  377, 412, 444, 474, 501, 527, 552, INT_MAX };
132 static const int16_t iquant_tbl40[] =
133  { INT16_MIN, -66, 28, 104, 169, 224, 274, 318,
134  358, 395, 429, 459, 488, 514, 539, 566,
135  566, 539, 514, 488, 459, 429, 395, 358,
136  318, 274, 224, 169, 104, 28, -66, INT16_MIN };
137 static const int16_t W_tbl40[] =
138  { 14, 14, 24, 39, 40, 41, 58, 100,
139  141, 179, 219, 280, 358, 440, 529, 696,
140  696, 529, 440, 358, 280, 219, 179, 141,
141  100, 58, 41, 40, 39, 24, 14, 14 };
142 static const uint8_t F_tbl40[] =
143  { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 3, 4, 5, 6, 6,
144  6, 6, 5, 4, 3, 2, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
145 
146 static const G726Tables G726Tables_pool[] =
151 
152 
156 static inline uint8_t quant(G726Context* c, int d)
157 {
158  int sign, exp, i, dln;
159 
160  sign = i = 0;
161  if (d < 0) {
162  sign = 1;
163  d = -d;
164  }
165  exp = av_log2_16bit(d);
166  dln = ((exp<<7) + (((d<<7)>>exp)&0x7f)) - (c->y>>2);
167 
168  while (c->tbls.quant[i] < INT_MAX && c->tbls.quant[i] < dln)
169  ++i;
170 
171  if (sign)
172  i = ~i;
173  if (c->code_size != 2 && i == 0) /* I'm not sure this is a good idea */
174  i = 0xff;
175 
176  return i;
177 }
178 
182 static inline int16_t inverse_quant(G726Context* c, int i)
183 {
184  int dql, dex, dqt;
185 
186  dql = c->tbls.iquant[i] + (c->y >> 2);
187  dex = (dql>>7) & 0xf; /* 4bit exponent */
188  dqt = (1<<7) + (dql & 0x7f); /* log2 -> linear */
189  return (dql < 0) ? 0 : ((dqt<<dex) >> 7);
190 }
191 
192 static int16_t g726_decode(G726Context* c, int I)
193 {
194  int dq, re_signal, pk0, fa1, i, tr, ylint, ylfrac, thr2, al, dq0;
195  Float11 f;
196  int I_sig= I >> (c->code_size - 1);
197 
198  dq = inverse_quant(c, I);
199 
200  /* Transition detect */
201  ylint = (c->yl >> 15);
202  ylfrac = (c->yl >> 10) & 0x1f;
203  thr2 = (ylint > 9) ? 0x1f << 10 : (0x20 + ylfrac) << ylint;
204  tr= (c->td == 1 && dq > ((3*thr2)>>2));
205 
206  if (I_sig) /* get the sign */
207  dq = -dq;
208  re_signal = c->se + dq;
209 
210  /* Update second order predictor coefficient A2 and A1 */
211  pk0 = (c->sez + dq) ? sgn(c->sez + dq) : 0;
212  dq0 = dq ? sgn(dq) : 0;
213  if (tr) {
214  c->a[0] = 0;
215  c->a[1] = 0;
216  for (i=0; i<6; i++)
217  c->b[i] = 0;
218  } else {
219  /* This is a bit crazy, but it really is +255 not +256 */
220  fa1 = av_clip((-c->a[0]*c->pk[0]*pk0)>>5, -256, 255);
221 
222  c->a[1] += 128*pk0*c->pk[1] + fa1 - (c->a[1]>>7);
223  c->a[1] = av_clip(c->a[1], -12288, 12288);
224  c->a[0] += 64*3*pk0*c->pk[0] - (c->a[0] >> 8);
225  c->a[0] = av_clip(c->a[0], -(15360 - c->a[1]), 15360 - c->a[1]);
226 
227  for (i=0; i<6; i++)
228  c->b[i] += 128*dq0*sgn(-c->dq[i].sign) - (c->b[i]>>8);
229  }
230 
231  /* Update Dq and Sr and Pk */
232  c->pk[1] = c->pk[0];
233  c->pk[0] = pk0 ? pk0 : 1;
234  c->sr[1] = c->sr[0];
235  i2f(re_signal, &c->sr[0]);
236  for (i=5; i>0; i--)
237  c->dq[i] = c->dq[i-1];
238  i2f(dq, &c->dq[0]);
239  c->dq[0].sign = I_sig; /* Isn't it crazy ?!?! */
240 
241  c->td = c->a[1] < -11776;
242 
243  /* Update Ap */
244  c->dms += (c->tbls.F[I]<<4) + ((- c->dms) >> 5);
245  c->dml += (c->tbls.F[I]<<4) + ((- c->dml) >> 7);
246  if (tr)
247  c->ap = 256;
248  else {
249  c->ap += (-c->ap) >> 4;
250  if (c->y <= 1535 || c->td || abs((c->dms << 2) - c->dml) >= (c->dml >> 3))
251  c->ap += 0x20;
252  }
253 
254  /* Update Yu and Yl */
255  c->yu = av_clip(c->y + c->tbls.W[I] + ((-c->y)>>5), 544, 5120);
256  c->yl += c->yu + ((-c->yl)>>6);
257 
258  /* Next iteration for Y */
259  al = (c->ap >= 256) ? 1<<6 : c->ap >> 2;
260  c->y = (c->yl + (c->yu - (c->yl>>6))*al) >> 6;
261 
262  /* Next iteration for SE and SEZ */
263  c->se = 0;
264  for (i=0; i<6; i++)
265  c->se += mult(i2f(c->b[i] >> 2, &f), &c->dq[i]);
266  c->sez = c->se >> 1;
267  for (i=0; i<2; i++)
268  c->se += mult(i2f(c->a[i] >> 2, &f), &c->sr[i]);
269  c->se >>= 1;
270 
271  return av_clip(re_signal << 2, -0xffff, 0xffff);
272 }
273 
275 {
276  int i;
277 
278  c->tbls = G726Tables_pool[c->code_size - 2];
279  for (i=0; i<2; i++) {
280  c->sr[i].mant = 1<<5;
281  c->pk[i] = 1;
282  }
283  for (i=0; i<6; i++) {
284  c->dq[i].mant = 1<<5;
285  }
286  c->yu = 544;
287  c->yl = 34816;
288 
289  c->y = 544;
290 
291  return 0;
292 }
293 
294 #if CONFIG_ADPCM_G726_ENCODER
295 static int16_t g726_encode(G726Context* c, int16_t sig)
296 {
297  uint8_t i;
298 
299  i = quant(c, sig/4 - c->se) & ((1<<c->code_size) - 1);
300  g726_decode(c, i);
301  return i;
302 }
303 
304 /* Interfacing to the libavcodec */
305 
306 static av_cold int g726_encode_init(AVCodecContext *avctx)
307 {
308  G726Context* c = avctx->priv_data;
309 
311  avctx->sample_rate != 8000) {
312  av_log(avctx, AV_LOG_ERROR, "Sample rates other than 8kHz are not "
313  "allowed when the compliance level is higher than unofficial. "
314  "Resample or reduce the compliance level.\n");
315  return AVERROR(EINVAL);
316  }
317  av_assert0(avctx->sample_rate > 0);
318 
319  if(avctx->channels != 1){
320  av_log(avctx, AV_LOG_ERROR, "Only mono is supported\n");
321  return AVERROR(EINVAL);
322  }
323 
324  if (avctx->bit_rate)
325  c->code_size = (avctx->bit_rate + avctx->sample_rate/2) / avctx->sample_rate;
326 
327  c->code_size = av_clip(c->code_size, 2, 5);
328  avctx->bit_rate = c->code_size * avctx->sample_rate;
329  avctx->bits_per_coded_sample = c->code_size;
330 
331  g726_reset(c);
332 
333  avctx->coded_frame = avcodec_alloc_frame();
334  if (!avctx->coded_frame)
335  return AVERROR(ENOMEM);
336  avctx->coded_frame->key_frame = 1;
337 
338  /* select a frame size that will end on a byte boundary and have a size of
339  approximately 1024 bytes */
340  avctx->frame_size = ((int[]){ 4096, 2736, 2048, 1640 })[c->code_size - 2];
341 
342  return 0;
343 }
344 
345 static av_cold int g726_encode_close(AVCodecContext *avctx)
346 {
347  av_freep(&avctx->coded_frame);
348  return 0;
349 }
350 
351 static int g726_encode_frame(AVCodecContext *avctx,
352  uint8_t *dst, int buf_size, void *data)
353 {
354  G726Context *c = avctx->priv_data;
355  const int16_t *samples = data;
356  PutBitContext pb;
357  int i;
358 
359  init_put_bits(&pb, dst, 1024*1024);
360 
361  for (i = 0; i < avctx->frame_size; i++)
362  put_bits(&pb, c->code_size, g726_encode(c, *samples++));
363 
364  flush_put_bits(&pb);
365 
366  return put_bits_count(&pb)>>3;
367 }
368 
369 #define OFFSET(x) offsetof(G726Context, x)
370 #define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
371 static const AVOption options[] = {
372  { "code_size", "Bits per code", OFFSET(code_size), AV_OPT_TYPE_INT, { 4 }, 2, 5, AE },
373  { NULL },
374 };
375 
376 static const AVClass class = {
377  .class_name = "g726",
378  .item_name = av_default_item_name,
379  .option = options,
380  .version = LIBAVUTIL_VERSION_INT,
381 };
382 
383 static const AVCodecDefault defaults[] = {
384  { "b", "0" },
385  { NULL },
386 };
387 
388 AVCodec ff_adpcm_g726_encoder = {
389  .name = "g726",
390  .type = AVMEDIA_TYPE_AUDIO,
391  .id = CODEC_ID_ADPCM_G726,
392  .priv_data_size = sizeof(G726Context),
393  .init = g726_encode_init,
394  .encode = g726_encode_frame,
395  .close = g726_encode_close,
396  .capabilities = CODEC_CAP_SMALL_LAST_FRAME,
397  .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
398  .long_name = NULL_IF_CONFIG_SMALL("G.726 ADPCM"),
399  .priv_class = &class,
400  .defaults = defaults,
401 };
402 #endif
403 
404 #if CONFIG_ADPCM_G726_DECODER
405 static av_cold int g726_decode_init(AVCodecContext *avctx)
406 {
407  G726Context* c = avctx->priv_data;
408 
410  avctx->sample_rate != 8000) {
411  av_log(avctx, AV_LOG_ERROR, "Only 8kHz sample rate is allowed when "
412  "the compliance level is strict. Reduce the compliance level "
413  "if you wish to decode the stream anyway.\n");
414  return AVERROR(EINVAL);
415  }
416 
417  if(avctx->channels != 1){
418  av_log(avctx, AV_LOG_ERROR, "Only mono is supported\n");
419  return AVERROR(EINVAL);
420  }
421 
422  c->code_size = avctx->bits_per_coded_sample;
423  if (c->code_size < 2 || c->code_size > 5) {
424  av_log(avctx, AV_LOG_ERROR, "Invalid number of bits %d\n", c->code_size);
425  return AVERROR(EINVAL);
426  }
427  g726_reset(c);
428 
429  avctx->sample_fmt = AV_SAMPLE_FMT_S16;
430 
432  avctx->coded_frame = &c->frame;
433 
434  return 0;
435 }
436 
437 static int g726_decode_frame(AVCodecContext *avctx, void *data,
438  int *got_frame_ptr, AVPacket *avpkt)
439 {
440  const uint8_t *buf = avpkt->data;
441  int buf_size = avpkt->size;
442  G726Context *c = avctx->priv_data;
443  int16_t *samples;
444  GetBitContext gb;
445  int out_samples, ret;
446 
447  out_samples = buf_size * 8 / c->code_size;
448 
449  /* get output buffer */
450  c->frame.nb_samples = out_samples;
451  if ((ret = ff_get_buffer(avctx, &c->frame)) < 0) {
452  av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
453  return ret;
454  }
455  samples = (int16_t *)c->frame.data[0];
456 
457  init_get_bits(&gb, buf, buf_size * 8);
458 
459  while (out_samples--)
460  *samples++ = g726_decode(c, get_bits(&gb, c->code_size));
461 
462  if (get_bits_left(&gb) > 0)
463  av_log(avctx, AV_LOG_ERROR, "Frame invalidly split, missing parser?\n");
464 
465  *got_frame_ptr = 1;
466  *(AVFrame *)data = c->frame;
467 
468  return buf_size;
469 }
470 
471 static void g726_decode_flush(AVCodecContext *avctx)
472 {
473  G726Context *c = avctx->priv_data;
474  g726_reset(c);
475 }
476 
477 AVCodec ff_adpcm_g726_decoder = {
478  .name = "g726",
479  .type = AVMEDIA_TYPE_AUDIO,
480  .id = CODEC_ID_ADPCM_G726,
481  .priv_data_size = sizeof(G726Context),
482  .init = g726_decode_init,
483  .decode = g726_decode_frame,
484  .flush = g726_decode_flush,
485  .capabilities = CODEC_CAP_DR1,
486  .long_name = NULL_IF_CONFIG_SMALL("G.726 ADPCM"),
487 };
488 #endif