Line data Source code
1 : /******************************************************************************************************
2 :
3 : (C) 2022-2025 IVAS codec Public Collaboration with portions copyright Dolby International AB, Ericsson AB,
4 : Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V., Huawei Technologies Co. LTD.,
5 : Koninklijke Philips N.V., Nippon Telegraph and Telephone Corporation, Nokia Technologies Oy, Orange,
6 : Panasonic Holdings Corporation, Qualcomm Technologies, Inc., VoiceAge Corporation, and other
7 : contributors to this repository. All Rights Reserved.
8 :
9 : This software is protected by copyright law and by international treaties.
10 : The IVAS codec Public Collaboration consisting of Dolby International AB, Ericsson AB,
11 : Fraunhofer-Gesellschaft zur Foerderung der angewandten Forschung e.V., Huawei Technologies Co. LTD.,
12 : Koninklijke Philips N.V., Nippon Telegraph and Telephone Corporation, Nokia Technologies Oy, Orange,
13 : Panasonic Holdings Corporation, Qualcomm Technologies, Inc., VoiceAge Corporation, and other
14 : contributors to this repository retain full ownership rights in their respective contributions in
15 : the software. This notice grants no license of any kind, including but not limited to patent
16 : license, nor is any license granted by implication, estoppel or otherwise.
17 :
18 : Contributors are required to enter into the IVAS codec Public Collaboration agreement before making
19 : contributions.
20 :
21 : This software is provided "AS IS", without any express or implied warranties. The software is in the
22 : development stage. It is intended exclusively for experts who have experience with such software and
23 : solely for the purpose of inspection. All implied warranties of non-infringement, merchantability
24 : and fitness for a particular purpose are hereby disclaimed and excluded.
25 :
26 : Any dispute, controversy or claim arising under or in relation to providing this software shall be
27 : submitted to and settled by the final, binding jurisdiction of the courts of Munich, Germany in
28 : accordance with the laws of the Federal Republic of Germany excluding its conflict of law rules and
29 : the United Nations Convention on Contracts on the International Sales of Goods.
30 :
31 : *******************************************************************************************************/
32 :
33 : #include <assert.h>
34 : #include <stdint.h>
35 : #include "options.h"
36 : #include <math.h>
37 : #include "ivas_cnst.h"
38 : #include "ivas_prot.h"
39 : #include "prot.h"
40 : #include "wmc_auto.h"
41 :
42 :
43 : /*-------------------------------------------------------------------*
44 : * Local function prototypes
45 : *-------------------------------------------------------------------*/
46 :
47 : static void inverseBwMS( const int16_t startLine, const int16_t stopLine, float x0[], float x1[], const float norm_fac );
48 :
49 :
50 : /*-------------------------------------------------------------------*
51 : * parse_stereo_from_bitstream
52 : *
53 : *
54 : *-------------------------------------------------------------------*/
55 :
56 2767283 : void parse_stereo_from_bitstream(
57 : STEREO_MDCT_DEC_DATA_HANDLE hStereoMdct, /* i/o: MDCT stereo decoder structure */
58 : Decoder_State **sts, /* i/o: decoder state structure */
59 : const int16_t mct_on, /* i : flag mct block (1) or stereo (0)*/
60 : const int16_t isSBAStereoMode, /* i : flag core coding for SBA */
61 : Decoder_State *st0, /* i/o: decoder state structure for Bstr*/
62 : int16_t ms_mask[NB_DIV][MAX_SFB] /* o : bandwise MS mask */
63 : )
64 : {
65 : int16_t i, k, nSubframes, mdct_stereo_mode;
66 : STEREO_MDCT_BAND_PARAMETERS *sfbConf;
67 : #ifdef DEBUGGING
68 : int16_t nbits_start = st0->next_bit_pos;
69 : #endif
70 :
71 2767283 : if ( !isSBAStereoMode )
72 : {
73 2172154 : nSubframes = ( sts[0]->core == TCX_10_CORE || ( sts[0]->core != sts[1]->core ) ) ? NB_DIV : 1;
74 2172154 : sfbConf = ( sts[0]->core == TCX_20_CORE ) ? &hStereoMdct->stbParamsTCX20 : &hStereoMdct->stbParamsTCX10;
75 2172154 : if ( sts[0]->last_core_from_bs == ACELP_CORE )
76 : {
77 6742 : sfbConf = &hStereoMdct->stbParamsTCX20afterACELP;
78 : }
79 :
80 2172154 : if ( hStereoMdct->use_itd )
81 : {
82 : int16_t I;
83 :
84 50538 : hStereoMdct->itd_mode = get_next_indice( st0, STEREO_DFT_ITD_MODE_NBITS );
85 : /*(*nb_bits) += STEREO_DFT_ITD_MODE_NBITS;*/ /*ITD mode flag: 1bit*/
86 :
87 50538 : hStereoMdct->itd = 0.f;
88 50538 : if ( hStereoMdct->itd_mode )
89 : {
90 22215 : /*(*nb_bits) += */ read_itd( st0, &I );
91 22215 : stereo_dft_dequantize_itd( &I, &hStereoMdct->itd, st0->output_Fs );
92 : }
93 : }
94 :
95 4413273 : for ( k = 0; k < nSubframes; k++ )
96 : {
97 2241119 : mdct_stereo_mode = get_next_indice( st0, 1 );
98 2241119 : if ( mdct_stereo_mode )
99 : {
100 2205298 : mdct_stereo_mode = 1 + get_next_indice( st0, 1 );
101 : }
102 2241119 : switch ( mdct_stereo_mode )
103 : {
104 35821 : case 0:
105 35821 : hStereoMdct->mdct_stereo_mode[k] = SMDCT_DUAL_MONO;
106 35821 : break;
107 1329815 : case 1:
108 1329815 : hStereoMdct->mdct_stereo_mode[k] = SMDCT_MS_FULL;
109 1329815 : break;
110 875483 : case 2:
111 875483 : hStereoMdct->mdct_stereo_mode[k] = SMDCT_BW_MS;
112 875483 : break;
113 0 : default:
114 0 : assert( !"Not supported stereo mode\n" );
115 : }
116 :
117 2241119 : if ( !mct_on )
118 : {
119 743703 : if ( sts[0]->core == sts[1]->core || k == 0 )
120 : {
121 740542 : hStereoMdct->global_ild[k] = get_next_indice( st0, SMDCT_GLOBAL_ILD_BITS );
122 1481084 : assert( ( hStereoMdct->global_ild[k] > 0 ) && ( hStereoMdct->global_ild[k] < SMDCT_ILD_RANGE ) );
123 : }
124 : else
125 : {
126 3161 : hStereoMdct->global_ild[1] = hStereoMdct->global_ild[0];
127 : }
128 : }
129 :
130 2241119 : set_s( ms_mask[k], ( hStereoMdct->mdct_stereo_mode[k] == SMDCT_MS_FULL ) ? 1 : 0, sfbConf->nBandsStereoCore );
131 :
132 2241119 : if ( hStereoMdct->mdct_stereo_mode[k] == SMDCT_BW_MS )
133 : {
134 39006202 : for ( i = 0; i < sfbConf->nBandsStereoCore; i++ )
135 : {
136 38130719 : ms_mask[k][i] = get_next_indice( st0, 1 );
137 : }
138 : }
139 :
140 2241119 : if ( st0->igf )
141 : {
142 1144601 : mdct_stereo_mode = get_next_indice( st0, 1 );
143 1144601 : if ( mdct_stereo_mode )
144 : {
145 674000 : mdct_stereo_mode = 1 + get_next_indice( st0, 1 );
146 : }
147 :
148 1144601 : switch ( mdct_stereo_mode )
149 : {
150 470601 : case 0:
151 470601 : hStereoMdct->IGFStereoMode[k] = SMDCT_DUAL_MONO;
152 470601 : break;
153 526855 : case 1:
154 526855 : hStereoMdct->IGFStereoMode[k] = SMDCT_MS_FULL;
155 526855 : break;
156 147145 : case 2:
157 147145 : hStereoMdct->IGFStereoMode[k] = SMDCT_BW_MS;
158 147145 : break;
159 0 : default:
160 0 : assert( !"Not supported stereo mode\n" );
161 : }
162 :
163 1144601 : set_s( &ms_mask[k][sfbConf->nBandsStereoCore], ( hStereoMdct->IGFStereoMode[k] == SMDCT_MS_FULL ) ? 1 : 0, sfbConf->sfbCnt - sfbConf->nBandsStereoCore );
164 :
165 1144601 : if ( hStereoMdct->IGFStereoMode[k] == SMDCT_BW_MS )
166 : {
167 1005917 : for ( i = sfbConf->nBandsStereoCore; i < sfbConf->sfbCnt; i++ )
168 : {
169 858772 : ms_mask[k][i] = get_next_indice( st0, 1 );
170 : }
171 : }
172 : }
173 : else
174 : {
175 1096518 : hStereoMdct->IGFStereoMode[k] = SMDCT_DUAL_MONO;
176 : }
177 : }
178 : }
179 :
180 2767283 : if ( !mct_on )
181 : {
182 1310339 : hStereoMdct->split_ratio = SMDCT_EQUAL_RATIO_RANGE; /* Equal bits to both channels */
183 1310339 : hStereoMdct->split_ratio = get_next_indice( st0, SMDCT_NBBITS_SPLIT_RATIO );
184 :
185 1310339 : assert( hStereoMdct->split_ratio > 0 );
186 : }
187 :
188 : #ifdef DEBUGGING
189 : {
190 : int16_t nBitsStereo = st0->next_bit_pos - nbits_start;
191 :
192 : assert( nBitsStereo <= st0->bits_frame_nominal );
193 : }
194 : #endif
195 :
196 2767283 : return;
197 : }
198 :
199 :
200 : /*-------------------------------------------------------------------*
201 : * stereo_decoder_tcx()
202 : *
203 : * apply stereo processing (inverse MS and global ILD)
204 : *-------------------------------------------------------------------*/
205 :
206 : #define NF_RED_FAC 0.75f
207 :
208 2277518 : void stereo_decoder_tcx(
209 : STEREO_MDCT_DEC_DATA *hStereoMdct, /* i/o: MDCT stereo decoder structure */
210 : int16_t ms_mask[NB_DIV][MAX_SFB], /* i : bandwise MS mask */
211 : float *spec_r_0[NB_DIV], /* i/o: spectrum right channel */
212 : float *spec_l[], /* i/o: spectrum left channel [NB_DIV][N] */
213 : float *spec_r[], /* i/o: spectrum right channel [NB_DIV][N] */
214 : const int16_t mdct_stereo_mode[], /* i : stereo mode (FB/band wise MS, dual mono */
215 : const int16_t core_l, /* i : core for left channel (TCX20/TCX10) */
216 : const int16_t core_r, /* i : core for right channel (TCX20/TCX10) */
217 : const int16_t igf, /* i : flag for IGF activity */
218 : const int16_t L_frameTCX_l, /* i : TCX frame length of left channel */
219 : const int16_t L_frameTCX_r, /* i : TCX frame length of right channel */
220 : const int16_t mct_on, /* i : flag mct block (1) or stereo (0) */
221 : const int16_t last_core_l, /* i : last core for left channel */
222 : const int16_t last_core_r, /* i : last core for right channel */
223 : const int16_t tmp_plc_upmix /* i : indicates temp upmix for PLC decision */
224 : )
225 : {
226 : int16_t i, k, sfb, nSubframes;
227 2277518 : STEREO_MDCT_BAND_PARAMETERS *sfbConf = NULL;
228 : float nrgRatio;
229 :
230 2277518 : nSubframes = 2;
231 2277518 : if ( ( core_l <= TCX_20_CORE && core_r <= TCX_20_CORE ) || tmp_plc_upmix )
232 : {
233 2208029 : nSubframes = 1;
234 : }
235 :
236 4624525 : for ( k = 0; k < nSubframes; k++ )
237 : {
238 2347007 : sfbConf = ( core_l == TCX_20_CORE ) ? &hStereoMdct->stbParamsTCX20 : &hStereoMdct->stbParamsTCX10;
239 :
240 2347007 : if ( last_core_l == ACELP_CORE || last_core_r == ACELP_CORE )
241 : {
242 6834 : sfbConf = &hStereoMdct->stbParamsTCX20afterACELP;
243 : }
244 :
245 2347007 : if ( mdct_stereo_mode[k] == SMDCT_MS_FULL )
246 : {
247 641733817 : for ( i = 0; i < sfbConf->sfbOffset[sfbConf->nBandsStereoCore]; i++ )
248 : {
249 640340294 : if ( spec_r_0[k][i] == 0.0f )
250 : {
251 242673733 : spec_r[k][i] *= NF_RED_FAC;
252 : }
253 : }
254 1393523 : inverseMS( sfbConf->sfbOffset[sfbConf->nBandsStereoCore], spec_l[k], spec_r[k], SQRT2_OVER_2 );
255 : }
256 953484 : else if ( mdct_stereo_mode[k] == SMDCT_BW_MS )
257 : {
258 40970961 : for ( sfb = 0; sfb < sfbConf->nBandsStereoCore; sfb++ )
259 : {
260 40053830 : if ( ms_mask[k][sfb] )
261 : {
262 314943256 : for ( i = sfbConf->sfbOffset[sfb]; i < sfbConf->sfbOffset[sfb + 1]; i++ )
263 : {
264 290935456 : if ( spec_r_0[k][i] == 0.0f )
265 : {
266 75864788 : spec_r[k][i] *= NF_RED_FAC;
267 : }
268 : }
269 24007800 : inverseBwMS( sfbConf->sfbOffset[sfb], sfbConf->sfbOffset[sfb + 1], spec_l[k], spec_r[k], SQRT2_OVER_2 );
270 : }
271 : }
272 : }
273 : #ifdef DEBUGGING
274 : else if ( mdct_stereo_mode[k] != SMDCT_DUAL_MONO )
275 : {
276 : assert( !"Not supported MDCT stereo mode!\n" );
277 : }
278 : #endif
279 :
280 2347007 : if ( igf )
281 : {
282 1188761 : if ( hStereoMdct->IGFStereoMode[k] == SMDCT_MS_FULL )
283 : {
284 167200775 : for ( i = sfbConf->sfbOffset[sfbConf->nBandsStereoCore]; i < sfbConf->sfbOffset[sfbConf->sfbCnt]; i++ )
285 : {
286 166654202 : if ( spec_r_0[k][i] == 0.0f )
287 : {
288 54353254 : spec_r[k][i] *= NF_RED_FAC;
289 : }
290 : }
291 546573 : inverseMS( sfbConf->sfbOffset[sfbConf->sfbCnt] - sfbConf->sfbOffset[sfbConf->nBandsStereoCore], &spec_l[k][sfbConf->sfbOffset[sfbConf->nBandsStereoCore]], &spec_r[k][sfbConf->sfbOffset[sfbConf->nBandsStereoCore]], SQRT2_OVER_2 );
292 : }
293 642188 : else if ( hStereoMdct->IGFStereoMode[k] == SMDCT_BW_MS )
294 : {
295 1028253 : for ( sfb = sfbConf->nBandsStereoCore; sfb < sfbConf->sfbCnt; sfb++ )
296 : {
297 878325 : if ( ms_mask[k][sfb] )
298 : {
299 22451499 : for ( i = sfbConf->sfbOffset[sfb]; i < sfbConf->sfbOffset[sfb + 1]; i++ )
300 : {
301 21984182 : if ( spec_r_0[k][i] == 0.0f )
302 : {
303 6179160 : spec_r[k][i] *= NF_RED_FAC;
304 : }
305 : }
306 467317 : inverseBwMS( sfbConf->sfbOffset[sfb], sfbConf->sfbOffset[sfb + 1], spec_l[k], spec_r[k], SQRT2_OVER_2 );
307 : }
308 : }
309 : }
310 : #ifdef DEBUGGING
311 : else if ( hStereoMdct->IGFStereoMode[k] != SMDCT_DUAL_MONO )
312 : {
313 : assert( !"Not supported MDCT stereo mode!\n" );
314 : }
315 : #endif
316 : }
317 :
318 2347007 : if ( !mct_on )
319 : {
320 849591 : nrgRatio = (float) SMDCT_ILD_RANGE / hStereoMdct->global_ild[k] - 1; /* nrgRatio = nrg[1]/nrg[0] */
321 849591 : hStereoMdct->smooth_ratio = 0.8f * hStereoMdct->smooth_ratio + 0.2f * nrgRatio;
322 : /* set flag to reverse dmx computation in case of right-side panning, only relevant for mono output */
323 849591 : if ( hStereoMdct->smooth_ratio > 1.3f )
324 : {
325 98304 : hStereoMdct->reverse_dmx = 1;
326 : }
327 751287 : else if ( hStereoMdct->smooth_ratio < 0.9f )
328 : {
329 113961 : hStereoMdct->reverse_dmx = 0;
330 : }
331 :
332 849591 : if ( ( nrgRatio > 1.0f ) && ( k < ( ( core_r == TCX_10_CORE ) ? NB_DIV : 1 ) ) )
333 : {
334 279224 : v_multc( spec_r[k], nrgRatio, spec_r[k], L_frameTCX_r );
335 : }
336 570367 : else if ( ( nrgRatio < 1.0f ) && ( k < ( ( core_l == TCX_10_CORE ) ? NB_DIV : 1 ) ) )
337 : {
338 152126 : v_multc( spec_l[k], 1.0f / nrgRatio, spec_l[k], L_frameTCX_l );
339 : }
340 : }
341 : } /* for k */
342 :
343 2277518 : return;
344 : }
345 :
346 :
347 : /*-------------------------------------------------------------------*
348 : * inverseBwMS()
349 : *
350 : * Band-wise M/S stereo processing
351 : *-------------------------------------------------------------------*/
352 :
353 26872852 : static void inverseBwMS(
354 : const int16_t startLine, /* i : start line of sfb */
355 : const int16_t stopLine, /* i : stop line of sfb */
356 : float x0[], /* i/o: mid/left channel coefficients */
357 : float x1[], /* i/o: side/right channel coefficients */
358 : const float norm_fac /* i : normalization factor */
359 : )
360 : {
361 : int16_t j;
362 : float tmpValue;
363 :
364 1154118166 : for ( j = startLine; j < stopLine; j++ )
365 : {
366 1127245314 : tmpValue = x0[j];
367 1127245314 : x0[j] = ( x0[j] + x1[j] ) * norm_fac;
368 1127245314 : x1[j] = ( tmpValue - x1[j] ) * norm_fac;
369 : }
370 :
371 26872852 : return;
372 : }
373 :
374 :
375 : /*-------------------------------------------------------------------*
376 : * inverseMS()
377 : *
378 : * M/S stereo processing
379 : *-------------------------------------------------------------------*/
380 :
381 2397735 : void inverseMS(
382 : const int16_t L_frame, /* i : frame length */
383 : float x0[], /* i/o: mid/left channel coefficients */
384 : float x1[], /* i/o: side/right channel coefficients */
385 : const float norm_fac /* i : normalization factor */
386 : )
387 : {
388 2397735 : inverseBwMS( 0, L_frame, x0, x1, norm_fac );
389 :
390 2397735 : return;
391 : }
392 :
393 :
394 : /*-------------------------------------------------------------------*
395 : * initMdctStereoDecData()
396 : *
397 : * Initialize MDCT stereo decoder configuration
398 : *-------------------------------------------------------------------*/
399 :
400 5285865 : void initMdctStereoDecData(
401 : STEREO_MDCT_DEC_DATA *hStereoMdct, /* i/o: mdct stereo parameters structure */
402 : const int16_t igf, /* i : flag indicating IGF activity */
403 : const H_IGF_GRID igfGrid, /* i : IGF grid configuration */
404 : const int32_t element_brate, /* i : element bitrate */
405 : const int16_t bwidth /* i : audio bandwidth */
406 : )
407 : {
408 : int16_t tcx_coded_lines;
409 :
410 5285865 : tcx_coded_lines = getNumTcxCodedLines( bwidth );
411 :
412 : /*Initialize sfb parameteres for TCX20 */
413 5285865 : stereo_mdct_init_bands( tcx_coded_lines, TCX_20_CORE, element_brate, igf, &igfGrid[IGF_GRID_LB_NORM], &hStereoMdct->stbParamsTCX20.sfbOffset[0], &hStereoMdct->stbParamsTCX20.sfbCnt );
414 :
415 : /*Initialize sfb parameteres for TCX10 */
416 5285865 : stereo_mdct_init_bands( tcx_coded_lines, TCX_10_CORE, element_brate, igf, &igfGrid[IGF_GRID_LB_SHORT], &hStereoMdct->stbParamsTCX10.sfbOffset[0], &hStereoMdct->stbParamsTCX10.sfbCnt );
417 :
418 : /*Initialize sfb parameteres for transition frames */
419 5285865 : stereo_mdct_init_bands( tcx_coded_lines, -1, element_brate, igf, &igfGrid[IGF_GRID_LB_TRAN], &hStereoMdct->stbParamsTCX20afterACELP.sfbOffset[0], &hStereoMdct->stbParamsTCX20afterACELP.sfbCnt );
420 :
421 5285865 : if ( igf )
422 : {
423 : /* calculate the igf start band from the igf start line */
424 2198025 : stereo_mdct_init_igf_start_band( &( hStereoMdct->stbParamsTCX20 ), 1.0f, bwidth, element_brate );
425 2198025 : stereo_mdct_init_igf_start_band( &( hStereoMdct->stbParamsTCX10 ), 0.5f, bwidth, element_brate );
426 2198025 : stereo_mdct_init_igf_start_band( &( hStereoMdct->stbParamsTCX20afterACELP ), 1.25f, bwidth, element_brate );
427 : }
428 : else
429 : {
430 3087840 : hStereoMdct->stbParamsTCX20.sfbIgfStart = -1;
431 3087840 : hStereoMdct->stbParamsTCX10.sfbIgfStart = -1;
432 3087840 : hStereoMdct->stbParamsTCX20afterACELP.sfbIgfStart = -1;
433 3087840 : hStereoMdct->stbParamsTCX10.nBandsStereoCore = hStereoMdct->stbParamsTCX10.sfbCnt;
434 3087840 : hStereoMdct->stbParamsTCX20.nBandsStereoCore = hStereoMdct->stbParamsTCX20.sfbCnt;
435 3087840 : hStereoMdct->stbParamsTCX20afterACELP.nBandsStereoCore = hStereoMdct->stbParamsTCX20afterACELP.sfbCnt;
436 : }
437 :
438 5285865 : return;
439 : }
440 :
441 :
442 : /*-------------------------------------------------------------------*
443 : * initMdctStereoDtxData()
444 : *
445 : * Allocate and initialize structures for MDCT-Stereo DTX operation
446 : *-------------------------------------------------------------------*/
447 :
448 89 : ivas_error initMdctStereoDtxData(
449 : CPE_DEC_HANDLE hCPE /* i/o: CPE handle */
450 : )
451 : {
452 : int16_t ch;
453 : ivas_error error;
454 :
455 89 : error = IVAS_ERR_OK;
456 :
457 267 : for ( ch = 0; ch < CPE_CHANNELS; ch++ )
458 : {
459 178 : DEC_CORE_HANDLE st = hCPE->hCoreCoder[ch];
460 :
461 178 : if ( st->hFdCngDec == NULL )
462 : {
463 : /* Create FD_CNG instance */
464 0 : if ( ( error = createFdCngDec( &st->hFdCngDec ) ) != IVAS_ERR_OK )
465 : {
466 0 : return error;
467 : }
468 :
469 : /* Init FD-CNG */
470 0 : initFdCngDec( st );
471 : }
472 :
473 178 : if ( st->first_CNG == 0 )
474 : {
475 178 : if ( ch == 1 && st->cng_sba_flag )
476 : {
477 49 : st->hFdCngDec->hFdCngCom->seed += 3;
478 : }
479 : }
480 :
481 178 : if ( st->cldfbAna == NULL )
482 : {
483 : /* open analysis for max. sampling rate 48kHz */
484 154 : if ( ( error = openCldfb( &st->cldfbAna, CLDFB_ANALYSIS, 48000, CLDFB_PROTOTYPE_1_25MS ) ) != IVAS_ERR_OK )
485 : {
486 0 : return error;
487 : }
488 : }
489 :
490 178 : if ( st->cldfbBPF == NULL )
491 : {
492 : /* open analysis BPF for max. internal sampling rate 16kHz */
493 154 : if ( ( error = openCldfb( &st->cldfbBPF, CLDFB_ANALYSIS, 16000, CLDFB_PROTOTYPE_1_25MS ) ) != IVAS_ERR_OK )
494 : {
495 0 : return error;
496 : }
497 : }
498 : }
499 :
500 89 : return error;
501 : }
502 :
503 : /*-------------------------------------------------------------------*
504 : * synchonize_channels_mdct_sid()
505 : *
506 : * Synchronize channels in SID frame in MDCT stereo
507 : *-------------------------------------------------------------------*/
508 :
509 16693809 : void synchonize_channels_mdct_sid(
510 : Decoder_State *sts[CPE_CHANNELS], /* i/o: decoder state structure */
511 : const int16_t n /* i : channel number */
512 : )
513 : {
514 : Decoder_State *st;
515 :
516 16693809 : st = sts[n];
517 :
518 16693809 : if ( st->element_mode == IVAS_CPE_MDCT && st->total_brate == SID_2k40 )
519 : {
520 3328 : if ( n == 1 )
521 : {
522 : /* synchronize channels */
523 1664 : sts[1]->L_frame = sts[0]->L_frame;
524 1664 : sts[1]->cng_type = sts[0]->cng_type;
525 1664 : sts[1]->bwidth = sts[0]->bwidth;
526 : #ifdef NONBE_MDCT_ST_DTX_FIX_SUBOPT_SPATIAL_CNG
527 1664 : sts[0]->hFdCngDec->hFdCngCom->coherence[0] = sts[1]->hFdCngDec->hFdCngCom->coherence[0]; /* coherence is stored in sts[1] - see ivas_decision_matrix_dec() and FdCngDecodeMDCTStereoSID() */
528 : #else
529 : sts[0]->hFdCngDec->hFdCngCom->coherence = sts[1]->hFdCngDec->hFdCngCom->coherence; /* coherence is stored in sts[1] - see ivas_decision_matrix_dec() */
530 : #endif
531 1664 : sts[0]->hFdCngDec->hFdCngCom->no_side_flag = sts[1]->hFdCngDec->hFdCngCom->no_side_flag;
532 :
533 : /* configure when there is a switching from DFT CNG to MDCT CNG */
534 1664 : if ( sts[0]->first_CNG == 1 && sts[1]->first_CNG == 0 )
535 : {
536 0 : configureFdCngDec( st->hFdCngDec, st->bwidth, st->element_brate, st->L_frame, st->last_L_frame, st->element_mode );
537 : }
538 : }
539 :
540 3328 : if ( sts[0]->first_CNG == 0 )
541 : {
542 : /* configure CNG after reading first side info from SID to get correct values for L_frame and bwidth if first SID is also first valid frame */
543 178 : configureFdCngDec( st->hFdCngDec, st->bwidth, st->element_brate, st->L_frame, st->last_L_frame, st->element_mode );
544 : }
545 : }
546 :
547 16693809 : return;
548 : }
549 :
550 :
551 : /*-------------------------------------------------------------------*
552 : * updateBuffersForDmxMdctStereo()
553 : *
554 : * synch buffers between channels for mono output and
555 : * apply passive downmix to certain buffers to enable smooth transitions
556 : * between active/inactive coding in MDCT-Stereo DTX
557 : *-------------------------------------------------------------------*/
558 :
559 4095 : void updateBuffersForDmxMdctStereo(
560 : CPE_DEC_HANDLE hCPE, /* i/o: CPE handle */
561 : const int16_t output_frame, /* i : output frame length */
562 : float *output[CPE_CHANNELS], /* i/o: decoder output */
563 : float synth[CPE_CHANNELS][L_FRAME48k] /* i/o: decoder synthesis */
564 : )
565 : {
566 : int16_t delay_buf_out_len, tcxltp_mem_in_len, delta, i;
567 : Decoder_State *sts[CPE_CHANNELS];
568 :
569 4095 : sts[0] = hCPE->hCoreCoder[0];
570 4095 : sts[1] = hCPE->hCoreCoder[1];
571 :
572 : /* synch buffers for inactive frames, but not for transition frames */
573 4095 : if ( hCPE->last_element_brate <= IVAS_SID_5k2 )
574 : {
575 3952 : mvr2r( output[0], output[1], output_frame );
576 3952 : mvr2r( synth[0], synth[1], output_frame );
577 : }
578 :
579 4095 : if ( hCPE->element_brate == IVAS_SID_5k2 && hCPE->last_element_brate > IVAS_SID_5k2 )
580 : {
581 : /* in the first SID frame after an active frame, create mid noise shape here, in SID frames that follow inactive frames, it is done directly in the SID decoding since the mid shape is being used in CNG then */
582 3494 : for ( int16_t p = 0; p < sts[0]->hFdCngDec->hFdCngCom->npart; p++ )
583 : {
584 3351 : sts[0]->hFdCngDec->hFdCngCom->sidNoiseEst[p] = 0.5f * ( sts[0]->hFdCngDec->hFdCngCom->sidNoiseEst[p] + sts[1]->hFdCngDec->hFdCngCom->sidNoiseEst[p] );
585 : }
586 : }
587 :
588 : /* for transition of active->inactive frame, apply passive downmix on buffers */
589 4095 : if ( hCPE->last_element_brate <= IVAS_SID_5k2 )
590 : {
591 3952 : delta = 1;
592 3952 : if ( output_frame == L_FRAME16k )
593 : {
594 1044 : delta = 2;
595 : }
596 2908 : else if ( output_frame == L_FRAME32k )
597 : {
598 1074 : delta = 4;
599 : }
600 1834 : else if ( output_frame == L_FRAME48k )
601 : {
602 1834 : delta = 6;
603 : }
604 :
605 3952 : delay_buf_out_len = delta * HQ_DELAY_COMP;
606 3952 : tcxltp_mem_in_len = NS2SA( sts[0]->output_Fs, TCXLTP_DELAY_NS );
607 :
608 3952 : assert( delay_buf_out_len > tcxltp_mem_in_len );
609 :
610 38728 : for ( i = 0; i < tcxltp_mem_in_len; i++ )
611 : {
612 34776 : sts[0]->hTcxLtpDec->tcxltp_mem_in[i] = INV_SQRT2 * ( sts[0]->hTcxLtpDec->tcxltp_mem_in[i] + sts[1]->hTcxLtpDec->tcxltp_mem_in[i] );
613 34776 : sts[0]->delay_buf_out[i] = INV_SQRT2 * ( sts[0]->delay_buf_out[i] + sts[1]->delay_buf_out[i] );
614 34776 : sts[0]->hHQ_core->old_out[i] = INV_SQRT2 * ( sts[0]->hHQ_core->old_out[i] + sts[1]->hHQ_core->old_out[i] );
615 34776 : sts[0]->hTcxLtpDec->tcxltp_mem_out[i] = INV_SQRT2 * ( sts[0]->hTcxLtpDec->tcxltp_mem_out[i] + sts[1]->hTcxLtpDec->tcxltp_mem_out[i] );
616 : }
617 143056 : for ( ; i < delay_buf_out_len; i++ )
618 : {
619 139104 : sts[0]->delay_buf_out[i] = INV_SQRT2 * ( sts[0]->delay_buf_out[i] + sts[1]->delay_buf_out[i] );
620 139104 : sts[0]->hHQ_core->old_out[i] = INV_SQRT2 * ( sts[0]->hHQ_core->old_out[i] + sts[1]->hHQ_core->old_out[i] );
621 139104 : sts[0]->hTcxLtpDec->tcxltp_mem_out[i] = INV_SQRT2 * ( sts[0]->hTcxLtpDec->tcxltp_mem_out[i] + sts[1]->hTcxLtpDec->tcxltp_mem_out[i] );
622 : }
623 2612152 : for ( ; i < output_frame; i++ )
624 : {
625 2608200 : sts[0]->hHQ_core->old_out[i] = INV_SQRT2 * ( sts[0]->hHQ_core->old_out[i] + sts[1]->hHQ_core->old_out[i] );
626 2608200 : sts[0]->hTcxLtpDec->tcxltp_mem_out[i] = INV_SQRT2 * ( sts[0]->hTcxLtpDec->tcxltp_mem_out[i] + sts[1]->hTcxLtpDec->tcxltp_mem_out[i] );
627 : }
628 : }
629 :
630 4095 : return;
631 : }
632 :
633 :
634 : /*-------------------------------------------------------------------*
635 : * applyDmxMdctStereo()
636 : *
637 : * apply passive downmix to certain buffers to enable smooth transitions
638 : * between active/inactive coding in MDCT-Stereo DTX
639 : *-------------------------------------------------------------------*/
640 :
641 48470 : void applyDmxMdctStereo(
642 : const CPE_DEC_HANDLE hCPE, /* i : CPE handle */
643 : float *output[CPE_CHANNELS], /* i/o: core decoder output */
644 : const int16_t output_frame /* i : output frame length */
645 : )
646 : {
647 : int16_t crossfade_len, i;
648 : int16_t dmx_len;
649 : float fade, step;
650 :
651 48470 : step = 1.f;
652 48470 : fade = 1.f;
653 48470 : dmx_len = output_frame;
654 :
655 48470 : if ( hCPE->last_element_brate <= IVAS_SID_5k2 )
656 : {
657 131 : crossfade_len = NS2SA( hCPE->hCoreCoder[0]->output_Fs, IVAS_DEC_DELAY_NS - DELAY_CLDFB_NS );
658 131 : step /= crossfade_len;
659 : }
660 : /* for first inactive CNG frame after active decoding we have to do a fade-OUT FROM the passive DMX */
661 48339 : else if ( hCPE->element_brate <= IVAS_SID_5k2 && hCPE->last_element_brate > IVAS_SID_5k2 )
662 : {
663 143 : crossfade_len = output_frame / 4;
664 143 : step /= -crossfade_len;
665 143 : fade = 0.f;
666 143 : dmx_len = crossfade_len;
667 : }
668 48196 : else if ( hCPE->last_element_mode == IVAS_CPE_DFT && hCPE->last_element_brate <= IVAS_32k )
669 : {
670 69 : crossfade_len = NS2SA( hCPE->hCoreCoder[0]->output_Fs, DELAY_CLDFB_NS );
671 69 : step /= crossfade_len;
672 : }
673 : else
674 : {
675 48127 : crossfade_len = 0;
676 : }
677 :
678 : /* apply crossfade */
679 88978 : for ( i = 0; i < crossfade_len; i++ )
680 : {
681 40508 : output[0][i] = output[0][i] * fade + ( output[0][i] + output[1][i] ) * INV_SQRT2 * ( 1 - fade );
682 40508 : fade -= step;
683 : }
684 :
685 : /* apply passive downmix on all-active-frame part */
686 34766362 : for ( ; i < dmx_len; i++ )
687 : {
688 34717892 : output[0][i] = ( output[0][i] + output[1][i] ) * INV_SQRT_2;
689 : }
690 :
691 48470 : return;
692 : }
|
