FFmpeg  4.0
g723_1enc.c
Go to the documentation of this file.
1 /*
2  * G.723.1 compatible encoder
3  * Copyright (c) Mohamed Naufal <naufal22@gmail.com>
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 /**
23  * @file
24  * G.723.1 compatible encoder
25  */
26 
27 #include <stdint.h>
28 #include <string.h>
29 
31 #include "libavutil/common.h"
32 #include "libavutil/mem.h"
33 #include "libavutil/opt.h"
34 
35 #include "avcodec.h"
36 #include "celp_math.h"
37 #include "g723_1.h"
38 #include "internal.h"
39 
40 #define BITSTREAM_WRITER_LE
41 #include "put_bits.h"
42 
44 {
45  G723_1_Context *p = avctx->priv_data;
46 
47  if (avctx->sample_rate != 8000) {
48  av_log(avctx, AV_LOG_ERROR, "Only 8000Hz sample rate supported\n");
49  return AVERROR(EINVAL);
50  }
51 
52  if (avctx->channels != 1) {
53  av_log(avctx, AV_LOG_ERROR, "Only mono supported\n");
54  return AVERROR(EINVAL);
55  }
56 
57  if (avctx->bit_rate == 6300) {
58  p->cur_rate = RATE_6300;
59  } else if (avctx->bit_rate == 5300) {
60  av_log(avctx, AV_LOG_ERROR, "Use bitrate 6300 instead of 5300.\n");
61  avpriv_report_missing_feature(avctx, "Bitrate 5300");
62  return AVERROR_PATCHWELCOME;
63  } else {
64  av_log(avctx, AV_LOG_ERROR, "Bitrate not supported, use 6300\n");
65  return AVERROR(EINVAL);
66  }
67  avctx->frame_size = 240;
68  memcpy(p->prev_lsp, dc_lsp, LPC_ORDER * sizeof(int16_t));
69 
70  return 0;
71 }
72 
73 /**
74  * Remove DC component from the input signal.
75  *
76  * @param buf input signal
77  * @param fir zero memory
78  * @param iir pole memory
79  */
80 static void highpass_filter(int16_t *buf, int16_t *fir, int *iir)
81 {
82  int i;
83  for (i = 0; i < FRAME_LEN; i++) {
84  *iir = (buf[i] << 15) + ((-*fir) << 15) + MULL2(*iir, 0x7f00);
85  *fir = buf[i];
86  buf[i] = av_clipl_int32((int64_t)*iir + (1 << 15)) >> 16;
87  }
88 }
89 
90 /**
91  * Estimate autocorrelation of the input vector.
92  *
93  * @param buf input buffer
94  * @param autocorr autocorrelation coefficients vector
95  */
96 static void comp_autocorr(int16_t *buf, int16_t *autocorr)
97 {
98  int i, scale, temp;
99  int16_t vector[LPC_FRAME];
100 
101  ff_g723_1_scale_vector(vector, buf, LPC_FRAME);
102 
103  /* Apply the Hamming window */
104  for (i = 0; i < LPC_FRAME; i++)
105  vector[i] = (vector[i] * hamming_window[i] + (1 << 14)) >> 15;
106 
107  /* Compute the first autocorrelation coefficient */
108  temp = ff_dot_product(vector, vector, LPC_FRAME);
109 
110  /* Apply a white noise correlation factor of (1025/1024) */
111  temp += temp >> 10;
112 
113  /* Normalize */
114  scale = ff_g723_1_normalize_bits(temp, 31);
115  autocorr[0] = av_clipl_int32((int64_t) (temp << scale) +
116  (1 << 15)) >> 16;
117 
118  /* Compute the remaining coefficients */
119  if (!autocorr[0]) {
120  memset(autocorr + 1, 0, LPC_ORDER * sizeof(int16_t));
121  } else {
122  for (i = 1; i <= LPC_ORDER; i++) {
123  temp = ff_dot_product(vector, vector + i, LPC_FRAME - i);
124  temp = MULL2((temp << scale), binomial_window[i - 1]);
125  autocorr[i] = av_clipl_int32((int64_t) temp + (1 << 15)) >> 16;
126  }
127  }
128 }
129 
130 /**
131  * Use Levinson-Durbin recursion to compute LPC coefficients from
132  * autocorrelation values.
133  *
134  * @param lpc LPC coefficients vector
135  * @param autocorr autocorrelation coefficients vector
136  * @param error prediction error
137  */
138 static void levinson_durbin(int16_t *lpc, int16_t *autocorr, int16_t error)
139 {
140  int16_t vector[LPC_ORDER];
141  int16_t partial_corr;
142  int i, j, temp;
143 
144  memset(lpc, 0, LPC_ORDER * sizeof(int16_t));
145 
146  for (i = 0; i < LPC_ORDER; i++) {
147  /* Compute the partial correlation coefficient */
148  temp = 0;
149  for (j = 0; j < i; j++)
150  temp -= lpc[j] * autocorr[i - j - 1];
151  temp = ((autocorr[i] << 13) + temp) << 3;
152 
153  if (FFABS(temp) >= (error << 16))
154  break;
155 
156  partial_corr = temp / (error << 1);
157 
158  lpc[i] = av_clipl_int32((int64_t) (partial_corr << 14) +
159  (1 << 15)) >> 16;
160 
161  /* Update the prediction error */
162  temp = MULL2(temp, partial_corr);
163  error = av_clipl_int32((int64_t) (error << 16) - temp +
164  (1 << 15)) >> 16;
165 
166  memcpy(vector, lpc, i * sizeof(int16_t));
167  for (j = 0; j < i; j++) {
168  temp = partial_corr * vector[i - j - 1] << 1;
169  lpc[j] = av_clipl_int32((int64_t) (lpc[j] << 16) - temp +
170  (1 << 15)) >> 16;
171  }
172  }
173 }
174 
175 /**
176  * Calculate LPC coefficients for the current frame.
177  *
178  * @param buf current frame
179  * @param prev_data 2 trailing subframes of the previous frame
180  * @param lpc LPC coefficients vector
181  */
182 static void comp_lpc_coeff(int16_t *buf, int16_t *lpc)
183 {
184  int16_t autocorr[(LPC_ORDER + 1) * SUBFRAMES];
185  int16_t *autocorr_ptr = autocorr;
186  int16_t *lpc_ptr = lpc;
187  int i, j;
188 
189  for (i = 0, j = 0; j < SUBFRAMES; i += SUBFRAME_LEN, j++) {
190  comp_autocorr(buf + i, autocorr_ptr);
191  levinson_durbin(lpc_ptr, autocorr_ptr + 1, autocorr_ptr[0]);
192 
193  lpc_ptr += LPC_ORDER;
194  autocorr_ptr += LPC_ORDER + 1;
195  }
196 }
197 
198 static void lpc2lsp(int16_t *lpc, int16_t *prev_lsp, int16_t *lsp)
199 {
200  int f[LPC_ORDER + 2]; ///< coefficients of the sum and difference
201  ///< polynomials (F1, F2) ordered as
202  ///< f1[0], f2[0], ...., f1[5], f2[5]
203 
204  int max, shift, cur_val, prev_val, count, p;
205  int i, j;
206  int64_t temp;
207 
208  /* Initialize f1[0] and f2[0] to 1 in Q25 */
209  for (i = 0; i < LPC_ORDER; i++)
210  lsp[i] = (lpc[i] * bandwidth_expand[i] + (1 << 14)) >> 15;
211 
212  /* Apply bandwidth expansion on the LPC coefficients */
213  f[0] = f[1] = 1 << 25;
214 
215  /* Compute the remaining coefficients */
216  for (i = 0; i < LPC_ORDER / 2; i++) {
217  /* f1 */
218  f[2 * i + 2] = -f[2 * i] - ((lsp[i] + lsp[LPC_ORDER - 1 - i]) << 12);
219  /* f2 */
220  f[2 * i + 3] = f[2 * i + 1] - ((lsp[i] - lsp[LPC_ORDER - 1 - i]) << 12);
221  }
222 
223  /* Divide f1[5] and f2[5] by 2 for use in polynomial evaluation */
224  f[LPC_ORDER] >>= 1;
225  f[LPC_ORDER + 1] >>= 1;
226 
227  /* Normalize and shorten */
228  max = FFABS(f[0]);
229  for (i = 1; i < LPC_ORDER + 2; i++)
230  max = FFMAX(max, FFABS(f[i]));
231 
232  shift = ff_g723_1_normalize_bits(max, 31);
233 
234  for (i = 0; i < LPC_ORDER + 2; i++)
235  f[i] = av_clipl_int32((int64_t) (f[i] << shift) + (1 << 15)) >> 16;
236 
237  /**
238  * Evaluate F1 and F2 at uniform intervals of pi/256 along the
239  * unit circle and check for zero crossings.
240  */
241  p = 0;
242  temp = 0;
243  for (i = 0; i <= LPC_ORDER / 2; i++)
244  temp += f[2 * i] * cos_tab[0];
245  prev_val = av_clipl_int32(temp << 1);
246  count = 0;
247  for (i = 1; i < COS_TBL_SIZE / 2; i++) {
248  /* Evaluate */
249  temp = 0;
250  for (j = 0; j <= LPC_ORDER / 2; j++)
251  temp += f[LPC_ORDER - 2 * j + p] * cos_tab[i * j % COS_TBL_SIZE];
252  cur_val = av_clipl_int32(temp << 1);
253 
254  /* Check for sign change, indicating a zero crossing */
255  if ((cur_val ^ prev_val) < 0) {
256  int abs_cur = FFABS(cur_val);
257  int abs_prev = FFABS(prev_val);
258  int sum = abs_cur + abs_prev;
259 
260  shift = ff_g723_1_normalize_bits(sum, 31);
261  sum <<= shift;
262  abs_prev = abs_prev << shift >> 8;
263  lsp[count++] = ((i - 1) << 7) + (abs_prev >> 1) / (sum >> 16);
264 
265  if (count == LPC_ORDER)
266  break;
267 
268  /* Switch between sum and difference polynomials */
269  p ^= 1;
270 
271  /* Evaluate */
272  temp = 0;
273  for (j = 0; j <= LPC_ORDER / 2; j++)
274  temp += f[LPC_ORDER - 2 * j + p] *
275  cos_tab[i * j % COS_TBL_SIZE];
276  cur_val = av_clipl_int32(temp << 1);
277  }
278  prev_val = cur_val;
279  }
280 
281  if (count != LPC_ORDER)
282  memcpy(lsp, prev_lsp, LPC_ORDER * sizeof(int16_t));
283 }
284 
285 /**
286  * Quantize the current LSP subvector.
287  *
288  * @param num band number
289  * @param offset offset of the current subvector in an LPC_ORDER vector
290  * @param size size of the current subvector
291  */
292 #define get_index(num, offset, size) \
293 { \
294  int error, max = -1; \
295  int16_t temp[4]; \
296  int i, j; \
297  \
298  for (i = 0; i < LSP_CB_SIZE; i++) { \
299  for (j = 0; j < size; j++){ \
300  temp[j] = (weight[j + (offset)] * lsp_band##num[i][j] + \
301  (1 << 14)) >> 15; \
302  } \
303  error = ff_g723_1_dot_product(lsp + (offset), temp, size) << 1; \
304  error -= ff_g723_1_dot_product(lsp_band##num[i], temp, size); \
305  if (error > max) { \
306  max = error; \
307  lsp_index[num] = i; \
308  } \
309  } \
310 }
311 
312 /**
313  * Vector quantize the LSP frequencies.
314  *
315  * @param lsp the current lsp vector
316  * @param prev_lsp the previous lsp vector
317  */
318 static void lsp_quantize(uint8_t *lsp_index, int16_t *lsp, int16_t *prev_lsp)
319 {
320  int16_t weight[LPC_ORDER];
321  int16_t min, max;
322  int shift, i;
323 
324  /* Calculate the VQ weighting vector */
325  weight[0] = (1 << 20) / (lsp[1] - lsp[0]);
326  weight[LPC_ORDER - 1] = (1 << 20) /
327  (lsp[LPC_ORDER - 1] - lsp[LPC_ORDER - 2]);
328 
329  for (i = 1; i < LPC_ORDER - 1; i++) {
330  min = FFMIN(lsp[i] - lsp[i - 1], lsp[i + 1] - lsp[i]);
331  if (min > 0x20)
332  weight[i] = (1 << 20) / min;
333  else
334  weight[i] = INT16_MAX;
335  }
336 
337  /* Normalize */
338  max = 0;
339  for (i = 0; i < LPC_ORDER; i++)
340  max = FFMAX(weight[i], max);
341 
342  shift = ff_g723_1_normalize_bits(max, 15);
343  for (i = 0; i < LPC_ORDER; i++) {
344  weight[i] <<= shift;
345  }
346 
347  /* Compute the VQ target vector */
348  for (i = 0; i < LPC_ORDER; i++) {
349  lsp[i] -= dc_lsp[i] +
350  (((prev_lsp[i] - dc_lsp[i]) * 12288 + (1 << 14)) >> 15);
351  }
352 
353  get_index(0, 0, 3);
354  get_index(1, 3, 3);
355  get_index(2, 6, 4);
356 }
357 
358 /**
359  * Perform IIR filtering.
360  *
361  * @param fir_coef FIR coefficients
362  * @param iir_coef IIR coefficients
363  * @param src source vector
364  * @param dest destination vector
365  */
366 static void iir_filter(int16_t *fir_coef, int16_t *iir_coef,
367  int16_t *src, int16_t *dest)
368 {
369  int m, n;
370 
371  for (m = 0; m < SUBFRAME_LEN; m++) {
372  int64_t filter = 0;
373  for (n = 1; n <= LPC_ORDER; n++) {
374  filter -= fir_coef[n - 1] * src[m - n] -
375  iir_coef[n - 1] * dest[m - n];
376  }
377 
378  dest[m] = av_clipl_int32((src[m] << 16) + (filter << 3) +
379  (1 << 15)) >> 16;
380  }
381 }
382 
383 /**
384  * Apply the formant perceptual weighting filter.
385  *
386  * @param flt_coef filter coefficients
387  * @param unq_lpc unquantized lpc vector
388  */
389 static void perceptual_filter(G723_1_Context *p, int16_t *flt_coef,
390  int16_t *unq_lpc, int16_t *buf)
391 {
392  int16_t vector[FRAME_LEN + LPC_ORDER];
393  int i, j, k, l = 0;
394 
395  memcpy(buf, p->iir_mem, sizeof(int16_t) * LPC_ORDER);
396  memcpy(vector, p->fir_mem, sizeof(int16_t) * LPC_ORDER);
397  memcpy(vector + LPC_ORDER, buf + LPC_ORDER, sizeof(int16_t) * FRAME_LEN);
398 
399  for (i = LPC_ORDER, j = 0; j < SUBFRAMES; i += SUBFRAME_LEN, j++) {
400  for (k = 0; k < LPC_ORDER; k++) {
401  flt_coef[k + 2 * l] = (unq_lpc[k + l] * percept_flt_tbl[0][k] +
402  (1 << 14)) >> 15;
403  flt_coef[k + 2 * l + LPC_ORDER] = (unq_lpc[k + l] *
404  percept_flt_tbl[1][k] +
405  (1 << 14)) >> 15;
406  }
407  iir_filter(flt_coef + 2 * l, flt_coef + 2 * l + LPC_ORDER,
408  vector + i, buf + i);
409  l += LPC_ORDER;
410  }
411  memcpy(p->iir_mem, buf + FRAME_LEN, sizeof(int16_t) * LPC_ORDER);
412  memcpy(p->fir_mem, vector + FRAME_LEN, sizeof(int16_t) * LPC_ORDER);
413 }
414 
415 /**
416  * Estimate the open loop pitch period.
417  *
418  * @param buf perceptually weighted speech
419  * @param start estimation is carried out from this position
420  */
421 static int estimate_pitch(int16_t *buf, int start)
422 {
423  int max_exp = 32;
424  int max_ccr = 0x4000;
425  int max_eng = 0x7fff;
426  int index = PITCH_MIN;
427  int offset = start - PITCH_MIN + 1;
428 
429  int ccr, eng, orig_eng, ccr_eng, exp;
430  int diff, temp;
431 
432  int i;
433 
434  orig_eng = ff_dot_product(buf + offset, buf + offset, HALF_FRAME_LEN);
435 
436  for (i = PITCH_MIN; i <= PITCH_MAX - 3; i++) {
437  offset--;
438 
439  /* Update energy and compute correlation */
440  orig_eng += buf[offset] * buf[offset] -
441  buf[offset + HALF_FRAME_LEN] * buf[offset + HALF_FRAME_LEN];
442  ccr = ff_dot_product(buf + start, buf + offset, HALF_FRAME_LEN);
443  if (ccr <= 0)
444  continue;
445 
446  /* Split into mantissa and exponent to maintain precision */
447  exp = ff_g723_1_normalize_bits(ccr, 31);
448  ccr = av_clipl_int32((int64_t) (ccr << exp) + (1 << 15)) >> 16;
449  exp <<= 1;
450  ccr *= ccr;
451  temp = ff_g723_1_normalize_bits(ccr, 31);
452  ccr = ccr << temp >> 16;
453  exp += temp;
454 
455  temp = ff_g723_1_normalize_bits(orig_eng, 31);
456  eng = av_clipl_int32((int64_t) (orig_eng << temp) + (1 << 15)) >> 16;
457  exp -= temp;
458 
459  if (ccr >= eng) {
460  exp--;
461  ccr >>= 1;
462  }
463  if (exp > max_exp)
464  continue;
465 
466  if (exp + 1 < max_exp)
467  goto update;
468 
469  /* Equalize exponents before comparison */
470  if (exp + 1 == max_exp)
471  temp = max_ccr >> 1;
472  else
473  temp = max_ccr;
474  ccr_eng = ccr * max_eng;
475  diff = ccr_eng - eng * temp;
476  if (diff > 0 && (i - index < PITCH_MIN || diff > ccr_eng >> 2)) {
477 update:
478  index = i;
479  max_exp = exp;
480  max_ccr = ccr;
481  max_eng = eng;
482  }
483  }
484  return index;
485 }
486 
487 /**
488  * Compute harmonic noise filter parameters.
489  *
490  * @param buf perceptually weighted speech
491  * @param pitch_lag open loop pitch period
492  * @param hf harmonic filter parameters
493  */
494 static void comp_harmonic_coeff(int16_t *buf, int16_t pitch_lag, HFParam *hf)
495 {
496  int ccr, eng, max_ccr, max_eng;
497  int exp, max, diff;
498  int energy[15];
499  int i, j;
500 
501  for (i = 0, j = pitch_lag - 3; j <= pitch_lag + 3; i++, j++) {
502  /* Compute residual energy */
503  energy[i << 1] = ff_dot_product(buf - j, buf - j, SUBFRAME_LEN);
504  /* Compute correlation */
505  energy[(i << 1) + 1] = ff_dot_product(buf, buf - j, SUBFRAME_LEN);
506  }
507 
508  /* Compute target energy */
509  energy[14] = ff_dot_product(buf, buf, SUBFRAME_LEN);
510 
511  /* Normalize */
512  max = 0;
513  for (i = 0; i < 15; i++)
514  max = FFMAX(max, FFABS(energy[i]));
515 
516  exp = ff_g723_1_normalize_bits(max, 31);
517  for (i = 0; i < 15; i++) {
518  energy[i] = av_clipl_int32((int64_t)(energy[i] << exp) +
519  (1 << 15)) >> 16;
520  }
521 
522  hf->index = -1;
523  hf->gain = 0;
524  max_ccr = 1;
525  max_eng = 0x7fff;
526 
527  for (i = 0; i <= 6; i++) {
528  eng = energy[i << 1];
529  ccr = energy[(i << 1) + 1];
530 
531  if (ccr <= 0)
532  continue;
533 
534  ccr = (ccr * ccr + (1 << 14)) >> 15;
535  diff = ccr * max_eng - eng * max_ccr;
536  if (diff > 0) {
537  max_ccr = ccr;
538  max_eng = eng;
539  hf->index = i;
540  }
541  }
542 
543  if (hf->index == -1) {
544  hf->index = pitch_lag;
545  return;
546  }
547 
548  eng = energy[14] * max_eng;
549  eng = (eng >> 2) + (eng >> 3);
550  ccr = energy[(hf->index << 1) + 1] * energy[(hf->index << 1) + 1];
551  if (eng < ccr) {
552  eng = energy[(hf->index << 1) + 1];
553 
554  if (eng >= max_eng)
555  hf->gain = 0x2800;
556  else
557  hf->gain = ((eng << 15) / max_eng * 0x2800 + (1 << 14)) >> 15;
558  }
559  hf->index += pitch_lag - 3;
560 }
561 
562 /**
563  * Apply the harmonic noise shaping filter.
564  *
565  * @param hf filter parameters
566  */
567 static void harmonic_filter(HFParam *hf, const int16_t *src, int16_t *dest)
568 {
569  int i;
570 
571  for (i = 0; i < SUBFRAME_LEN; i++) {
572  int64_t temp = hf->gain * src[i - hf->index] << 1;
573  dest[i] = av_clipl_int32((src[i] << 16) - temp + (1 << 15)) >> 16;
574  }
575 }
576 
577 static void harmonic_noise_sub(HFParam *hf, const int16_t *src, int16_t *dest)
578 {
579  int i;
580  for (i = 0; i < SUBFRAME_LEN; i++) {
581  int64_t temp = hf->gain * src[i - hf->index] << 1;
582  dest[i] = av_clipl_int32(((dest[i] - src[i]) << 16) + temp +
583  (1 << 15)) >> 16;
584  }
585 }
586 
587 /**
588  * Combined synthesis and formant perceptual weighting filer.
589  *
590  * @param qnt_lpc quantized lpc coefficients
591  * @param perf_lpc perceptual filter coefficients
592  * @param perf_fir perceptual filter fir memory
593  * @param perf_iir perceptual filter iir memory
594  * @param scale the filter output will be scaled by 2^scale
595  */
596 static void synth_percept_filter(int16_t *qnt_lpc, int16_t *perf_lpc,
597  int16_t *perf_fir, int16_t *perf_iir,
598  const int16_t *src, int16_t *dest, int scale)
599 {
600  int i, j;
601  int16_t buf_16[SUBFRAME_LEN + LPC_ORDER];
602  int64_t buf[SUBFRAME_LEN];
603 
604  int16_t *bptr_16 = buf_16 + LPC_ORDER;
605 
606  memcpy(buf_16, perf_fir, sizeof(int16_t) * LPC_ORDER);
607  memcpy(dest - LPC_ORDER, perf_iir, sizeof(int16_t) * LPC_ORDER);
608 
609  for (i = 0; i < SUBFRAME_LEN; i++) {
610  int64_t temp = 0;
611  for (j = 1; j <= LPC_ORDER; j++)
612  temp -= qnt_lpc[j - 1] * bptr_16[i - j];
613 
614  buf[i] = (src[i] << 15) + (temp << 3);
615  bptr_16[i] = av_clipl_int32(buf[i] + (1 << 15)) >> 16;
616  }
617 
618  for (i = 0; i < SUBFRAME_LEN; i++) {
619  int64_t fir = 0, iir = 0;
620  for (j = 1; j <= LPC_ORDER; j++) {
621  fir -= perf_lpc[j - 1] * bptr_16[i - j];
622  iir += perf_lpc[j + LPC_ORDER - 1] * dest[i - j];
623  }
624  dest[i] = av_clipl_int32(((buf[i] + (fir << 3)) << scale) + (iir << 3) +
625  (1 << 15)) >> 16;
626  }
627  memcpy(perf_fir, buf_16 + SUBFRAME_LEN, sizeof(int16_t) * LPC_ORDER);
628  memcpy(perf_iir, dest + SUBFRAME_LEN - LPC_ORDER,
629  sizeof(int16_t) * LPC_ORDER);
630 }
631 
632 /**
633  * Compute the adaptive codebook contribution.
634  *
635  * @param buf input signal
636  * @param index the current subframe index
637  */
638 static void acb_search(G723_1_Context *p, int16_t *residual,
639  int16_t *impulse_resp, const int16_t *buf,
640  int index)
641 {
642  int16_t flt_buf[PITCH_ORDER][SUBFRAME_LEN];
643 
644  const int16_t *cb_tbl = adaptive_cb_gain85;
645 
646  int ccr_buf[PITCH_ORDER * SUBFRAMES << 2];
647 
648  int pitch_lag = p->pitch_lag[index >> 1];
649  int acb_lag = 1;
650  int acb_gain = 0;
651  int odd_frame = index & 1;
652  int iter = 3 + odd_frame;
653  int count = 0;
654  int tbl_size = 85;
655 
656  int i, j, k, l, max;
657  int64_t temp;
658 
659  if (!odd_frame) {
660  if (pitch_lag == PITCH_MIN)
661  pitch_lag++;
662  else
663  pitch_lag = FFMIN(pitch_lag, PITCH_MAX - 5);
664  }
665 
666  for (i = 0; i < iter; i++) {
667  ff_g723_1_get_residual(residual, p->prev_excitation, pitch_lag + i - 1);
668 
669  for (j = 0; j < SUBFRAME_LEN; j++) {
670  temp = 0;
671  for (k = 0; k <= j; k++)
672  temp += residual[PITCH_ORDER - 1 + k] * impulse_resp[j - k];
673  flt_buf[PITCH_ORDER - 1][j] = av_clipl_int32((temp << 1) +
674  (1 << 15)) >> 16;
675  }
676 
677  for (j = PITCH_ORDER - 2; j >= 0; j--) {
678  flt_buf[j][0] = ((residual[j] << 13) + (1 << 14)) >> 15;
679  for (k = 1; k < SUBFRAME_LEN; k++) {
680  temp = (flt_buf[j + 1][k - 1] << 15) +
681  residual[j] * impulse_resp[k];
682  flt_buf[j][k] = av_clipl_int32((temp << 1) + (1 << 15)) >> 16;
683  }
684  }
685 
686  /* Compute crosscorrelation with the signal */
687  for (j = 0; j < PITCH_ORDER; j++) {
688  temp = ff_dot_product(buf, flt_buf[j], SUBFRAME_LEN);
689  ccr_buf[count++] = av_clipl_int32(temp << 1);
690  }
691 
692  /* Compute energies */
693  for (j = 0; j < PITCH_ORDER; j++) {
694  ccr_buf[count++] = ff_g723_1_dot_product(flt_buf[j], flt_buf[j],
695  SUBFRAME_LEN);
696  }
697 
698  for (j = 1; j < PITCH_ORDER; j++) {
699  for (k = 0; k < j; k++) {
700  temp = ff_dot_product(flt_buf[j], flt_buf[k], SUBFRAME_LEN);
701  ccr_buf[count++] = av_clipl_int32(temp << 2);
702  }
703  }
704  }
705 
706  /* Normalize and shorten */
707  max = 0;
708  for (i = 0; i < 20 * iter; i++)
709  max = FFMAX(max, FFABS(ccr_buf[i]));
710 
711  temp = ff_g723_1_normalize_bits(max, 31);
712 
713  for (i = 0; i < 20 * iter; i++)
714  ccr_buf[i] = av_clipl_int32((int64_t) (ccr_buf[i] << temp) +
715  (1 << 15)) >> 16;
716 
717  max = 0;
718  for (i = 0; i < iter; i++) {
719  /* Select quantization table */
720  if (!odd_frame && pitch_lag + i - 1 >= SUBFRAME_LEN - 2 ||
721  odd_frame && pitch_lag >= SUBFRAME_LEN - 2) {
722  cb_tbl = adaptive_cb_gain170;
723  tbl_size = 170;
724  }
725 
726  for (j = 0, k = 0; j < tbl_size; j++, k += 20) {
727  temp = 0;
728  for (l = 0; l < 20; l++)
729  temp += ccr_buf[20 * i + l] * cb_tbl[k + l];
730  temp = av_clipl_int32(temp);
731 
732  if (temp > max) {
733  max = temp;
734  acb_gain = j;
735  acb_lag = i;
736  }
737  }
738  }
739 
740  if (!odd_frame) {
741  pitch_lag += acb_lag - 1;
742  acb_lag = 1;
743  }
744 
745  p->pitch_lag[index >> 1] = pitch_lag;
746  p->subframe[index].ad_cb_lag = acb_lag;
747  p->subframe[index].ad_cb_gain = acb_gain;
748 }
749 
750 /**
751  * Subtract the adaptive codebook contribution from the input
752  * to obtain the residual.
753  *
754  * @param buf target vector
755  */
756 static void sub_acb_contrib(const int16_t *residual, const int16_t *impulse_resp,
757  int16_t *buf)
758 {
759  int i, j;
760  /* Subtract adaptive CB contribution to obtain the residual */
761  for (i = 0; i < SUBFRAME_LEN; i++) {
762  int64_t temp = buf[i] << 14;
763  for (j = 0; j <= i; j++)
764  temp -= residual[j] * impulse_resp[i - j];
765 
766  buf[i] = av_clipl_int32((temp << 2) + (1 << 15)) >> 16;
767  }
768 }
769 
770 /**
771  * Quantize the residual signal using the fixed codebook (MP-MLQ).
772  *
773  * @param optim optimized fixed codebook parameters
774  * @param buf excitation vector
775  */
776 static void get_fcb_param(FCBParam *optim, int16_t *impulse_resp,
777  int16_t *buf, int pulse_cnt, int pitch_lag)
778 {
779  FCBParam param;
780  int16_t impulse_r[SUBFRAME_LEN];
781  int16_t temp_corr[SUBFRAME_LEN];
782  int16_t impulse_corr[SUBFRAME_LEN];
783 
784  int ccr1[SUBFRAME_LEN];
785  int ccr2[SUBFRAME_LEN];
786  int amp, err, max, max_amp_index, min, scale, i, j, k, l;
787 
788  int64_t temp;
789 
790  /* Update impulse response */
791  memcpy(impulse_r, impulse_resp, sizeof(int16_t) * SUBFRAME_LEN);
792  param.dirac_train = 0;
793  if (pitch_lag < SUBFRAME_LEN - 2) {
794  param.dirac_train = 1;
795  ff_g723_1_gen_dirac_train(impulse_r, pitch_lag);
796  }
797 
798  for (i = 0; i < SUBFRAME_LEN; i++)
799  temp_corr[i] = impulse_r[i] >> 1;
800 
801  /* Compute impulse response autocorrelation */
802  temp = ff_g723_1_dot_product(temp_corr, temp_corr, SUBFRAME_LEN);
803 
804  scale = ff_g723_1_normalize_bits(temp, 31);
805  impulse_corr[0] = av_clipl_int32((temp << scale) + (1 << 15)) >> 16;
806 
807  for (i = 1; i < SUBFRAME_LEN; i++) {
808  temp = ff_g723_1_dot_product(temp_corr + i, temp_corr,
809  SUBFRAME_LEN - i);
810  impulse_corr[i] = av_clipl_int32((temp << scale) + (1 << 15)) >> 16;
811  }
812 
813  /* Compute crosscorrelation of impulse response with residual signal */
814  scale -= 4;
815  for (i = 0; i < SUBFRAME_LEN; i++) {
816  temp = ff_g723_1_dot_product(buf + i, impulse_r, SUBFRAME_LEN - i);
817  if (scale < 0)
818  ccr1[i] = temp >> -scale;
819  else
820  ccr1[i] = av_clipl_int32(temp << scale);
821  }
822 
823  /* Search loop */
824  for (i = 0; i < GRID_SIZE; i++) {
825  /* Maximize the crosscorrelation */
826  max = 0;
827  for (j = i; j < SUBFRAME_LEN; j += GRID_SIZE) {
828  temp = FFABS(ccr1[j]);
829  if (temp >= max) {
830  max = temp;
831  param.pulse_pos[0] = j;
832  }
833  }
834 
835  /* Quantize the gain (max crosscorrelation/impulse_corr[0]) */
836  amp = max;
837  min = 1 << 30;
838  max_amp_index = GAIN_LEVELS - 2;
839  for (j = max_amp_index; j >= 2; j--) {
840  temp = av_clipl_int32((int64_t) fixed_cb_gain[j] *
841  impulse_corr[0] << 1);
842  temp = FFABS(temp - amp);
843  if (temp < min) {
844  min = temp;
845  max_amp_index = j;
846  }
847  }
848 
849  max_amp_index--;
850  /* Select additional gain values */
851  for (j = 1; j < 5; j++) {
852  for (k = i; k < SUBFRAME_LEN; k += GRID_SIZE) {
853  temp_corr[k] = 0;
854  ccr2[k] = ccr1[k];
855  }
856  param.amp_index = max_amp_index + j - 2;
857  amp = fixed_cb_gain[param.amp_index];
858 
859  param.pulse_sign[0] = (ccr2[param.pulse_pos[0]] < 0) ? -amp : amp;
860  temp_corr[param.pulse_pos[0]] = 1;
861 
862  for (k = 1; k < pulse_cnt; k++) {
863  max = INT_MIN;
864  for (l = i; l < SUBFRAME_LEN; l += GRID_SIZE) {
865  if (temp_corr[l])
866  continue;
867  temp = impulse_corr[FFABS(l - param.pulse_pos[k - 1])];
868  temp = av_clipl_int32((int64_t) temp *
869  param.pulse_sign[k - 1] << 1);
870  ccr2[l] -= temp;
871  temp = FFABS(ccr2[l]);
872  if (temp > max) {
873  max = temp;
874  param.pulse_pos[k] = l;
875  }
876  }
877 
878  param.pulse_sign[k] = (ccr2[param.pulse_pos[k]] < 0) ?
879  -amp : amp;
880  temp_corr[param.pulse_pos[k]] = 1;
881  }
882 
883  /* Create the error vector */
884  memset(temp_corr, 0, sizeof(int16_t) * SUBFRAME_LEN);
885 
886  for (k = 0; k < pulse_cnt; k++)
887  temp_corr[param.pulse_pos[k]] = param.pulse_sign[k];
888 
889  for (k = SUBFRAME_LEN - 1; k >= 0; k--) {
890  temp = 0;
891  for (l = 0; l <= k; l++) {
892  int prod = av_clipl_int32((int64_t) temp_corr[l] *
893  impulse_r[k - l] << 1);
894  temp = av_clipl_int32(temp + prod);
895  }
896  temp_corr[k] = temp << 2 >> 16;
897  }
898 
899  /* Compute square of error */
900  err = 0;
901  for (k = 0; k < SUBFRAME_LEN; k++) {
902  int64_t prod;
903  prod = av_clipl_int32((int64_t) buf[k] * temp_corr[k] << 1);
904  err = av_clipl_int32(err - prod);
905  prod = av_clipl_int32((int64_t) temp_corr[k] * temp_corr[k]);
906  err = av_clipl_int32(err + prod);
907  }
908 
909  /* Minimize */
910  if (err < optim->min_err) {
911  optim->min_err = err;
912  optim->grid_index = i;
913  optim->amp_index = param.amp_index;
914  optim->dirac_train = param.dirac_train;
915 
916  for (k = 0; k < pulse_cnt; k++) {
917  optim->pulse_sign[k] = param.pulse_sign[k];
918  optim->pulse_pos[k] = param.pulse_pos[k];
919  }
920  }
921  }
922  }
923 }
924 
925 /**
926  * Encode the pulse position and gain of the current subframe.
927  *
928  * @param optim optimized fixed CB parameters
929  * @param buf excitation vector
930  */
931 static void pack_fcb_param(G723_1_Subframe *subfrm, FCBParam *optim,
932  int16_t *buf, int pulse_cnt)
933 {
934  int i, j;
935 
936  j = PULSE_MAX - pulse_cnt;
937 
938  subfrm->pulse_sign = 0;
939  subfrm->pulse_pos = 0;
940 
941  for (i = 0; i < SUBFRAME_LEN >> 1; i++) {
942  int val = buf[optim->grid_index + (i << 1)];
943  if (!val) {
944  subfrm->pulse_pos += combinatorial_table[j][i];
945  } else {
946  subfrm->pulse_sign <<= 1;
947  if (val < 0)
948  subfrm->pulse_sign++;
949  j++;
950 
951  if (j == PULSE_MAX)
952  break;
953  }
954  }
955  subfrm->amp_index = optim->amp_index;
956  subfrm->grid_index = optim->grid_index;
957  subfrm->dirac_train = optim->dirac_train;
958 }
959 
960 /**
961  * Compute the fixed codebook excitation.
962  *
963  * @param buf target vector
964  * @param impulse_resp impulse response of the combined filter
965  */
966 static void fcb_search(G723_1_Context *p, int16_t *impulse_resp,
967  int16_t *buf, int index)
968 {
969  FCBParam optim;
970  int pulse_cnt = pulses[index];
971  int i;
972 
973  optim.min_err = 1 << 30;
974  get_fcb_param(&optim, impulse_resp, buf, pulse_cnt, SUBFRAME_LEN);
975 
976  if (p->pitch_lag[index >> 1] < SUBFRAME_LEN - 2) {
977  get_fcb_param(&optim, impulse_resp, buf, pulse_cnt,
978  p->pitch_lag[index >> 1]);
979  }
980 
981  /* Reconstruct the excitation */
982  memset(buf, 0, sizeof(int16_t) * SUBFRAME_LEN);
983  for (i = 0; i < pulse_cnt; i++)
984  buf[optim.pulse_pos[i]] = optim.pulse_sign[i];
985 
986  pack_fcb_param(&p->subframe[index], &optim, buf, pulse_cnt);
987 
988  if (optim.dirac_train)
989  ff_g723_1_gen_dirac_train(buf, p->pitch_lag[index >> 1]);
990 }
991 
992 /**
993  * Pack the frame parameters into output bitstream.
994  *
995  * @param frame output buffer
996  * @param size size of the buffer
997  */
998 static int pack_bitstream(G723_1_Context *p, AVPacket *avpkt)
999 {
1000  PutBitContext pb;
1001  int info_bits = 0;
1002  int i, temp;
1003 
1004  init_put_bits(&pb, avpkt->data, avpkt->size);
1005 
1006  put_bits(&pb, 2, info_bits);
1007 
1008  put_bits(&pb, 8, p->lsp_index[2]);
1009  put_bits(&pb, 8, p->lsp_index[1]);
1010  put_bits(&pb, 8, p->lsp_index[0]);
1011 
1012  put_bits(&pb, 7, p->pitch_lag[0] - PITCH_MIN);
1013  put_bits(&pb, 2, p->subframe[1].ad_cb_lag);
1014  put_bits(&pb, 7, p->pitch_lag[1] - PITCH_MIN);
1015  put_bits(&pb, 2, p->subframe[3].ad_cb_lag);
1016 
1017  /* Write 12 bit combined gain */
1018  for (i = 0; i < SUBFRAMES; i++) {
1019  temp = p->subframe[i].ad_cb_gain * GAIN_LEVELS +
1020  p->subframe[i].amp_index;
1021  if (p->cur_rate == RATE_6300)
1022  temp += p->subframe[i].dirac_train << 11;
1023  put_bits(&pb, 12, temp);
1024  }
1025 
1026  put_bits(&pb, 1, p->subframe[0].grid_index);
1027  put_bits(&pb, 1, p->subframe[1].grid_index);
1028  put_bits(&pb, 1, p->subframe[2].grid_index);
1029  put_bits(&pb, 1, p->subframe[3].grid_index);
1030 
1031  if (p->cur_rate == RATE_6300) {
1032  skip_put_bits(&pb, 1); /* reserved bit */
1033 
1034  /* Write 13 bit combined position index */
1035  temp = (p->subframe[0].pulse_pos >> 16) * 810 +
1036  (p->subframe[1].pulse_pos >> 14) * 90 +
1037  (p->subframe[2].pulse_pos >> 16) * 9 +
1038  (p->subframe[3].pulse_pos >> 14);
1039  put_bits(&pb, 13, temp);
1040 
1041  put_bits(&pb, 16, p->subframe[0].pulse_pos & 0xffff);
1042  put_bits(&pb, 14, p->subframe[1].pulse_pos & 0x3fff);
1043  put_bits(&pb, 16, p->subframe[2].pulse_pos & 0xffff);
1044  put_bits(&pb, 14, p->subframe[3].pulse_pos & 0x3fff);
1045 
1046  put_bits(&pb, 6, p->subframe[0].pulse_sign);
1047  put_bits(&pb, 5, p->subframe[1].pulse_sign);
1048  put_bits(&pb, 6, p->subframe[2].pulse_sign);
1049  put_bits(&pb, 5, p->subframe[3].pulse_sign);
1050  }
1051 
1052  flush_put_bits(&pb);
1053  return frame_size[info_bits];
1054 }
1055 
1056 static int g723_1_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
1057  const AVFrame *frame, int *got_packet_ptr)
1058 {
1059  G723_1_Context *p = avctx->priv_data;
1060  int16_t unq_lpc[LPC_ORDER * SUBFRAMES];
1061  int16_t qnt_lpc[LPC_ORDER * SUBFRAMES];
1062  int16_t cur_lsp[LPC_ORDER];
1063  int16_t weighted_lpc[LPC_ORDER * SUBFRAMES << 1];
1064  int16_t vector[FRAME_LEN + PITCH_MAX];
1065  int offset, ret, i, j;
1066  int16_t *in, *start;
1067  HFParam hf[4];
1068 
1069  /* duplicate input */
1070  start = in = av_malloc(frame->nb_samples * sizeof(int16_t));
1071  if (!in)
1072  return AVERROR(ENOMEM);
1073  memcpy(in, frame->data[0], frame->nb_samples * sizeof(int16_t));
1074 
1075  highpass_filter(in, &p->hpf_fir_mem, &p->hpf_iir_mem);
1076 
1077  memcpy(vector, p->prev_data, HALF_FRAME_LEN * sizeof(int16_t));
1078  memcpy(vector + HALF_FRAME_LEN, in, FRAME_LEN * sizeof(int16_t));
1079 
1080  comp_lpc_coeff(vector, unq_lpc);
1081  lpc2lsp(&unq_lpc[LPC_ORDER * 3], p->prev_lsp, cur_lsp);
1082  lsp_quantize(p->lsp_index, cur_lsp, p->prev_lsp);
1083 
1084  /* Update memory */
1085  memcpy(vector + LPC_ORDER, p->prev_data + SUBFRAME_LEN,
1086  sizeof(int16_t) * SUBFRAME_LEN);
1087  memcpy(vector + LPC_ORDER + SUBFRAME_LEN, in,
1088  sizeof(int16_t) * (HALF_FRAME_LEN + SUBFRAME_LEN));
1089  memcpy(p->prev_data, in + HALF_FRAME_LEN,
1090  sizeof(int16_t) * HALF_FRAME_LEN);
1091  memcpy(in, vector + LPC_ORDER, sizeof(int16_t) * FRAME_LEN);
1092 
1093  perceptual_filter(p, weighted_lpc, unq_lpc, vector);
1094 
1095  memcpy(in, vector + LPC_ORDER, sizeof(int16_t) * FRAME_LEN);
1096  memcpy(vector, p->prev_weight_sig, sizeof(int16_t) * PITCH_MAX);
1097  memcpy(vector + PITCH_MAX, in, sizeof(int16_t) * FRAME_LEN);
1098 
1099  ff_g723_1_scale_vector(vector, vector, FRAME_LEN + PITCH_MAX);
1100 
1101  p->pitch_lag[0] = estimate_pitch(vector, PITCH_MAX);
1102  p->pitch_lag[1] = estimate_pitch(vector, PITCH_MAX + HALF_FRAME_LEN);
1103 
1104  for (i = PITCH_MAX, j = 0; j < SUBFRAMES; i += SUBFRAME_LEN, j++)
1105  comp_harmonic_coeff(vector + i, p->pitch_lag[j >> 1], hf + j);
1106 
1107  memcpy(vector, p->prev_weight_sig, sizeof(int16_t) * PITCH_MAX);
1108  memcpy(vector + PITCH_MAX, in, sizeof(int16_t) * FRAME_LEN);
1109  memcpy(p->prev_weight_sig, vector + FRAME_LEN, sizeof(int16_t) * PITCH_MAX);
1110 
1111  for (i = 0, j = 0; j < SUBFRAMES; i += SUBFRAME_LEN, j++)
1112  harmonic_filter(hf + j, vector + PITCH_MAX + i, in + i);
1113 
1114  ff_g723_1_inverse_quant(cur_lsp, p->prev_lsp, p->lsp_index, 0);
1115  ff_g723_1_lsp_interpolate(qnt_lpc, cur_lsp, p->prev_lsp);
1116 
1117  memcpy(p->prev_lsp, cur_lsp, sizeof(int16_t) * LPC_ORDER);
1118 
1119  offset = 0;
1120  for (i = 0; i < SUBFRAMES; i++) {
1121  int16_t impulse_resp[SUBFRAME_LEN];
1122  int16_t residual[SUBFRAME_LEN + PITCH_ORDER - 1];
1123  int16_t flt_in[SUBFRAME_LEN];
1124  int16_t zero[LPC_ORDER], fir[LPC_ORDER], iir[LPC_ORDER];
1125 
1126  /**
1127  * Compute the combined impulse response of the synthesis filter,
1128  * formant perceptual weighting filter and harmonic noise shaping filter
1129  */
1130  memset(zero, 0, sizeof(int16_t) * LPC_ORDER);
1131  memset(vector, 0, sizeof(int16_t) * PITCH_MAX);
1132  memset(flt_in, 0, sizeof(int16_t) * SUBFRAME_LEN);
1133 
1134  flt_in[0] = 1 << 13; /* Unit impulse */
1135  synth_percept_filter(qnt_lpc + offset, weighted_lpc + (offset << 1),
1136  zero, zero, flt_in, vector + PITCH_MAX, 1);
1137  harmonic_filter(hf + i, vector + PITCH_MAX, impulse_resp);
1138 
1139  /* Compute the combined zero input response */
1140  flt_in[0] = 0;
1141  memcpy(fir, p->perf_fir_mem, sizeof(int16_t) * LPC_ORDER);
1142  memcpy(iir, p->perf_iir_mem, sizeof(int16_t) * LPC_ORDER);
1143 
1144  synth_percept_filter(qnt_lpc + offset, weighted_lpc + (offset << 1),
1145  fir, iir, flt_in, vector + PITCH_MAX, 0);
1146  memcpy(vector, p->harmonic_mem, sizeof(int16_t) * PITCH_MAX);
1147  harmonic_noise_sub(hf + i, vector + PITCH_MAX, in);
1148 
1149  acb_search(p, residual, impulse_resp, in, i);
1151  p->pitch_lag[i >> 1], &p->subframe[i],
1152  p->cur_rate);
1153  sub_acb_contrib(residual, impulse_resp, in);
1154 
1155  fcb_search(p, impulse_resp, in, i);
1156 
1157  /* Reconstruct the excitation */
1159  p->pitch_lag[i >> 1], &p->subframe[i],
1160  RATE_6300);
1161 
1162  memmove(p->prev_excitation, p->prev_excitation + SUBFRAME_LEN,
1163  sizeof(int16_t) * (PITCH_MAX - SUBFRAME_LEN));
1164  for (j = 0; j < SUBFRAME_LEN; j++)
1165  in[j] = av_clip_int16((in[j] << 1) + impulse_resp[j]);
1166  memcpy(p->prev_excitation + PITCH_MAX - SUBFRAME_LEN, in,
1167  sizeof(int16_t) * SUBFRAME_LEN);
1168 
1169  /* Update filter memories */
1170  synth_percept_filter(qnt_lpc + offset, weighted_lpc + (offset << 1),
1171  p->perf_fir_mem, p->perf_iir_mem,
1172  in, vector + PITCH_MAX, 0);
1173  memmove(p->harmonic_mem, p->harmonic_mem + SUBFRAME_LEN,
1174  sizeof(int16_t) * (PITCH_MAX - SUBFRAME_LEN));
1175  memcpy(p->harmonic_mem + PITCH_MAX - SUBFRAME_LEN, vector + PITCH_MAX,
1176  sizeof(int16_t) * SUBFRAME_LEN);
1177 
1178  in += SUBFRAME_LEN;
1179  offset += LPC_ORDER;
1180  }
1181 
1182  av_free(start);
1183 
1184  if ((ret = ff_alloc_packet2(avctx, avpkt, 24, 0)) < 0)
1185  return ret;
1186 
1187  *got_packet_ptr = 1;
1188  avpkt->size = pack_bitstream(p, avpkt);
1189  return 0;
1190 }
1191 
1193  .name = "g723_1",
1194  .long_name = NULL_IF_CONFIG_SMALL("G.723.1"),
1195  .type = AVMEDIA_TYPE_AUDIO,
1196  .id = AV_CODEC_ID_G723_1,
1197  .priv_data_size = sizeof(G723_1_Context),
1199  .encode2 = g723_1_encode_frame,
1200  .sample_fmts = (const enum AVSampleFormat[]) {
1202  },
1203 };
const char const char void * val
Definition: avisynth_c.h:771
static void pack_fcb_param(G723_1_Subframe *subfrm, FCBParam *optim, int16_t *buf, int pulse_cnt)
Encode the pulse position and gain of the current subframe.
Definition: g723_1enc.c:931
#define COS_TBL_SIZE
Definition: g723_1.h:49
static void comp_autocorr(int16_t *buf, int16_t *autocorr)
Estimate autocorrelation of the input vector.
Definition: g723_1enc.c:96
static int shift(int a, int b)
Definition: sonic.c:82
int grid_index
Definition: g723_1.h:113
int dirac_train
Definition: g723_1.h:83
This structure describes decoded (raw) audio or video data.
Definition: frame.h:218
int ad_cb_gain
Definition: g723_1.h:82
int pitch_lag[2]
Definition: g723_1.h:127
int amp_index
Definition: g723_1.h:112
static void put_bits(Jpeg2000EncoderContext *s, int val, int n)
put n times val bit
Definition: j2kenc.c:207
int64_t bit_rate
the average bitrate
Definition: avcodec.h:1568
int16_t prev_weight_sig[PITCH_MAX]
Definition: g723_1.h:153
Memory handling functions.
else temp
Definition: vf_mcdeint.c:256
static av_cold int init(AVCodecContext *avctx)
Definition: avrndec.c:35
G723.1 unpacked data subframe.
Definition: g723_1.h:80
int16_t fir_mem[LPC_ORDER]
Definition: g723_1.h:135
static float cos_tab[256]
Definition: dca_lbr.c:123
static const int8_t pulses[4]
Number of non-zero pulses in the MP-MLQ excitation.
Definition: g723_1.h:720
int size
Definition: avcodec.h:1431
static void skip_put_bits(PutBitContext *s, int n)
Skip the given number of bits.
Definition: put_bits.h:346
#define PITCH_ORDER
Definition: g723_1.h:45
int min_err
Definition: g723_1.h:111
int index
Definition: g723_1.h:103
#define src
Definition: vp8dsp.c:254
AVCodec.
Definition: avcodec.h:3408
#define PITCH_MIN
Definition: g723_1.h:43
int pulse_pos[PULSE_MAX]
Definition: g723_1.h:115
#define FRAME_LEN
Definition: g723_1.h:37
int dirac_train
Definition: g723_1.h:114
int ff_alloc_packet2(AVCodecContext *avctx, AVPacket *avpkt, int64_t size, int64_t min_size)
Check AVPacket size and/or allocate data.
Definition: encode.c:32
void ff_g723_1_inverse_quant(int16_t *cur_lsp, int16_t *prev_lsp, uint8_t *lsp_index, int bad_frame)
Perform inverse quantization of LSP frequencies.
Definition: g723_1.c:201
static void filter(int16_t *output, ptrdiff_t out_stride, int16_t *low, ptrdiff_t low_stride, int16_t *high, ptrdiff_t high_stride, int len, int clip)
Definition: cfhd.c:114
static void sub_acb_contrib(const int16_t *residual, const int16_t *impulse_resp, int16_t *buf)
Subtract the adaptive codebook contribution from the input to obtain the residual.
Definition: g723_1enc.c:756
uint8_t
#define av_cold
Definition: attributes.h:82
#define av_malloc(s)
Optimized fixed codebook excitation parameters.
Definition: g723_1.h:110
AVOptions.
#define LPC_ORDER
Definition: g723_1.h:40
static const int16_t adaptive_cb_gain85[85 *20]
Definition: g723_1.h:733
static AVFrame * frame
int pulse_sign
Definition: g723_1.h:84
uint8_t * data
Definition: avcodec.h:1430
static const int16_t percept_flt_tbl[2][LPC_ORDER]
0.5^i scaled by 2^15
Definition: g723_1.h:1428
static void levinson_durbin(int16_t *lpc, int16_t *autocorr, int16_t error)
Use Levinson-Durbin recursion to compute LPC coefficients from autocorrelation values.
Definition: g723_1enc.c:138
void ff_g723_1_lsp_interpolate(int16_t *lpc, int16_t *cur_lsp, int16_t *prev_lsp)
Quantize LSP frequencies by interpolation and convert them to the corresponding LPC coefficients...
Definition: g723_1.c:180
#define GRID_SIZE
Definition: g723_1.h:46
#define av_log(a,...)
int16_t prev_data[HALF_FRAME_LEN]
Definition: g723_1.h:152
static const int16_t adaptive_cb_gain170[170 *20]
Definition: g723_1.h:949
static const int32_t combinatorial_table[PULSE_MAX][SUBFRAME_LEN/GRID_SIZE]
Used for the coding/decoding of the pulses positions for the MP-MLQ codebook.
Definition: g723_1.h:627
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:176
AVCodec ff_g723_1_encoder
Definition: g723_1enc.c:1192
static av_always_inline void update(SilenceDetectContext *s, AVFrame *insamples, int is_silence, int current_sample, int64_t nb_samples_notify, AVRational time_base)
int ff_g723_1_normalize_bits(int num, int width)
Calculate the number of left-shifts required for normalizing the input.
Definition: g723_1.c:49
#define AVERROR(e)
Definition: error.h:43
static int g723_1_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, const AVFrame *frame, int *got_packet_ptr)
Definition: g723_1enc.c:1056
int amp_index
Definition: g723_1.h:86
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification. ...
Definition: internal.h:186
#define LPC_FRAME
Definition: g723_1.h:39
void ff_g723_1_gen_dirac_train(int16_t *buf, int pitch_lag)
Generate a train of dirac functions with period as pitch lag.
Definition: g723_1.c:74
int pulse_sign[PULSE_MAX]
Definition: g723_1.h:116
#define zero
Definition: regdef.h:64
int grid_index
Definition: g723_1.h:85
const char * name
Name of the codec implementation.
Definition: avcodec.h:3415
int64_t ff_dot_product(const int16_t *a, const int16_t *b, int length)
Calculate the dot product of 2 int16_t vectors.
Definition: celp_math.c:98
int16_t prev_excitation[PITCH_MAX]
Definition: g723_1.h:132
static void harmonic_filter(HFParam *hf, const int16_t *src, int16_t *dest)
Apply the harmonic noise shaping filter.
Definition: g723_1enc.c:567
static const uint8_t offset[127][2]
Definition: vf_spp.c:92
static void lsp_quantize(uint8_t *lsp_index, int16_t *lsp, int16_t *prev_lsp)
Vector quantize the LSP frequencies.
Definition: g723_1enc.c:318
#define FFMAX(a, b)
Definition: common.h:94
int8_t exp
Definition: eval.c:72
uint64_t residual
Definition: dirac_vlc.h:29
void ff_g723_1_gen_acb_excitation(int16_t *vector, int16_t *prev_excitation, int pitch_lag, G723_1_Subframe *subfrm, enum Rate cur_rate)
Generate adaptive codebook excitation.
Definition: g723_1.c:86
G723_1_Subframe subframe[4]
Definition: g723_1.h:122
#define PITCH_MAX
Definition: g723_1.h:44
static const int16_t fixed_cb_gain[GAIN_LEVELS]
Definition: g723_1.h:727
static void fcb_search(G723_1_Context *p, int16_t *impulse_resp, int16_t *buf, int index)
Compute the fixed codebook excitation.
Definition: g723_1enc.c:966
static av_cold int g723_1_encode_init(AVCodecContext *avctx)
Definition: g723_1enc.c:43
enum Rate cur_rate
Definition: g723_1.h:125
static void harmonic_noise_sub(HFParam *hf, const int16_t *src, int16_t *dest)
Definition: g723_1enc.c:577
audio channel layout utility functions
#define FFMIN(a, b)
Definition: common.h:96
static int pack_bitstream(G723_1_Context *p, AVPacket *avpkt)
Pack the frame parameters into output bitstream.
Definition: g723_1enc.c:998
static void iir_filter(int16_t *fir_coef, int16_t *iir_coef, int16_t *src, int16_t *dest)
Perform IIR filtering.
Definition: g723_1enc.c:366
static void comp_lpc_coeff(int16_t *buf, int16_t *lpc)
Calculate LPC coefficients for the current frame.
Definition: g723_1enc.c:182
void ff_g723_1_get_residual(int16_t *residual, int16_t *prev_excitation, int lag)
Get delayed contribution from the previous excitation vector.
Definition: g723_1.c:60
int ff_g723_1_dot_product(const int16_t *a, const int16_t *b, int length)
Definition: g723_1.c:54
#define HALF_FRAME_LEN
Definition: g723_1.h:38
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:72
int n
Definition: avisynth_c.h:684
static void error(const char *err)
#define GAIN_LEVELS
Definition: g723_1.h:48
#define AVERROR_PATCHWELCOME
Not yet implemented in FFmpeg, patches welcome.
Definition: error.h:62
int frame_size
Number of samples per channel in an audio frame.
Definition: avcodec.h:2193
int frame_size
Definition: mxfenc.c:1947
int ff_g723_1_scale_vector(int16_t *dst, const int16_t *vector, int length)
Scale vector contents based on the largest of their absolutes.
Definition: g723_1.c:32
Libavcodec external API header.
static const int16_t dc_lsp[LPC_ORDER]
LSP DC component.
Definition: g723_1.h:229
AVSampleFormat
Audio sample formats.
Definition: samplefmt.h:58
int sample_rate
samples per second
Definition: avcodec.h:2173
main external API structure.
Definition: avcodec.h:1518
static void highpass_filter(int16_t *buf, int16_t *fir, int *iir)
Remove DC component from the input signal.
Definition: g723_1enc.c:80
G.723.1 types, functions and data tables.
void * buf
Definition: avisynth_c.h:690
uint8_t pi<< 24) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0f/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_U8, uint8_t,(*(const uint8_t *) pi - 0x80) *(1.0/(1<< 7))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S16, int16_t,(*(const int16_t *) pi >> 8)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0f/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S16, int16_t, *(const int16_t *) pi *(1.0/(1<< 15))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_S32, int32_t,(*(const int32_t *) pi >> 24)+0x80) CONV_FUNC_GROUP(AV_SAMPLE_FMT_FLT, float, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0f/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, double, AV_SAMPLE_FMT_S32, int32_t, *(const int32_t *) pi *(1.0/(1U<< 31))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_FLT, float, av_clip_uint8(lrintf(*(const float *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, float, av_clip_int16(lrintf(*(const float *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, float, av_clipl_int32(llrintf(*(const float *) pi *(1U<< 31)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_U8, uint8_t, AV_SAMPLE_FMT_DBL, double, av_clip_uint8(lrint(*(const double *) pi *(1<< 7))+0x80)) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, double, av_clip_int16(lrint(*(const double *) pi *(1<< 15)))) CONV_FUNC_GROUP(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, double, av_clipl_int32(llrint(*(const double *) pi *(1U<< 31)))) #define SET_CONV_FUNC_GROUP(ofmt, ifmt) static void set_generic_function(AudioConvert *ac) { } void ff_audio_convert_free(AudioConvert **ac) { if(! *ac) return;ff_dither_free(&(*ac) ->dc);av_freep(ac);} AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr, enum AVSampleFormat out_fmt, enum AVSampleFormat in_fmt, int channels, int sample_rate, int apply_map) { AudioConvert *ac;int in_planar, out_planar;ac=av_mallocz(sizeof(*ac));if(!ac) return NULL;ac->avr=avr;ac->out_fmt=out_fmt;ac->in_fmt=in_fmt;ac->channels=channels;ac->apply_map=apply_map;if(avr->dither_method !=AV_RESAMPLE_DITHER_NONE &&av_get_packed_sample_fmt(out_fmt)==AV_SAMPLE_FMT_S16 &&av_get_bytes_per_sample(in_fmt) > 2) { ac->dc=ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate, apply_map);if(!ac->dc) { av_free(ac);return NULL;} return ac;} in_planar=ff_sample_fmt_is_planar(in_fmt, channels);out_planar=ff_sample_fmt_is_planar(out_fmt, channels);if(in_planar==out_planar) { ac->func_type=CONV_FUNC_TYPE_FLAT;ac->planes=in_planar ? ac->channels :1;} else if(in_planar) ac->func_type=CONV_FUNC_TYPE_INTERLEAVE;else ac->func_type=CONV_FUNC_TYPE_DEINTERLEAVE;set_generic_function(ac);if(ARCH_AARCH64) ff_audio_convert_init_aarch64(ac);if(ARCH_ARM) ff_audio_convert_init_arm(ac);if(ARCH_X86) ff_audio_convert_init_x86(ac);return ac;} int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in) { int use_generic=1;int len=in->nb_samples;int p;if(ac->dc) { av_log(ac->avr, AV_LOG_TRACE, "%d samples - audio_convert: %s to %s (dithered)\", len, av_get_sample_fmt_name(ac->in_fmt), av_get_sample_fmt_name(ac->out_fmt));return ff_convert_dither(ac-> in
#define PULSE_MAX
Definition: dss_sp.c:32
static void perceptual_filter(G723_1_Context *p, int16_t *flt_coef, int16_t *unq_lpc, int16_t *buf)
Apply the formant perceptual weighting filter.
Definition: g723_1enc.c:389
static const int16_t hamming_window[LPC_FRAME]
Hamming window coefficients scaled by 2^15.
Definition: g723_1.h:1390
int index
Definition: gxfenc.c:89
int16_t harmonic_mem[PITCH_MAX]
Definition: g723_1.h:160
static void synth_percept_filter(int16_t *qnt_lpc, int16_t *perf_lpc, int16_t *perf_fir, int16_t *perf_iir, const int16_t *src, int16_t *dest, int scale)
Combined synthesis and formant perceptual weighting filer.
Definition: g723_1enc.c:596
int16_t hpf_fir_mem
highpass filter fir
Definition: g723_1.h:155
static int weight(int i, int blen, int offset)
Definition: diracdec.c:1523
#define SUBFRAME_LEN
Definition: g723_1.h:36
static void get_fcb_param(FCBParam *optim, int16_t *impulse_resp, int16_t *buf, int pulse_cnt, int pitch_lag)
Quantize the residual signal using the fixed codebook (MP-MLQ).
Definition: g723_1enc.c:776
#define get_index(num, offset, size)
Quantize the current LSP subvector.
Definition: g723_1enc.c:292
void avpriv_report_missing_feature(void *avc, const char *msg,...) av_printf_format(2
Log a generic warning message about a missing feature.
int16_t prev_lsp[LPC_ORDER]
Definition: g723_1.h:130
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:232
int hpf_iir_mem
and iir memories
Definition: g723_1.h:156
static void lpc2lsp(int16_t *lpc, int16_t *prev_lsp, int16_t *lsp)
Definition: g723_1enc.c:198
#define SUBFRAMES
Definition: dcaenc.c:50
common internal api header.
static void flush_put_bits(PutBitContext *s)
Pad the end of the output stream with zeros.
Definition: put_bits.h:101
common internal and external API header
int16_t perf_iir_mem[LPC_ORDER]
and iir memories
Definition: g723_1.h:158
signed 16 bits
Definition: samplefmt.h:61
static int estimate_pitch(int16_t *buf, int start)
Estimate the open loop pitch period.
Definition: g723_1enc.c:421
static void init_put_bits(PutBitContext *s, uint8_t *buffer, int buffer_size)
Initialize the PutBitContext s.
Definition: put_bits.h:48
Harmonic filter parameters.
Definition: g723_1.h:102
void * priv_data
Definition: avcodec.h:1545
static av_always_inline int diff(const uint32_t a, const uint32_t b)
#define av_free(p)
#define MULL2(a, b)
Bitexact implementation of 2ab scaled by 1/2^16.
Definition: g723_1.h:57
int channels
number of audio channels
Definition: avcodec.h:2174
static void comp_harmonic_coeff(int16_t *buf, int16_t pitch_lag, HFParam *hf)
Compute harmonic noise filter parameters.
Definition: g723_1enc.c:494
uint8_t lsp_index[LSP_BANDS]
Definition: g723_1.h:126
int pulse_pos
Definition: g723_1.h:87
static const int16_t bandwidth_expand[LPC_ORDER]
0.994^i scaled by 2^15
Definition: g723_1.h:1421
int iir_mem[LPC_ORDER]
Definition: g723_1.h:136
static enum AVSampleFormat sample_fmts[]
Definition: adpcmenc.c:701
void INT64 INT64 count
Definition: avisynth_c.h:690
void INT64 start
Definition: avisynth_c.h:690
int16_t perf_fir_mem[LPC_ORDER]
perceptual filter fir
Definition: g723_1.h:157
static void acb_search(G723_1_Context *p, int16_t *residual, int16_t *impulse_resp, const int16_t *buf, int index)
Compute the adaptive codebook contribution.
Definition: g723_1enc.c:638
float min
This structure stores compressed data.
Definition: avcodec.h:1407
int nb_samples
number of audio samples (per channel) described by this frame
Definition: frame.h:284
for(j=16;j >0;--j)
static const int16_t binomial_window[LPC_ORDER]
Binomial window coefficients scaled by 2^15.
Definition: g723_1.h:1414
int ad_cb_lag
adaptive codebook lag
Definition: g723_1.h:81
int gain
Definition: g723_1.h:104
bitstream writer API