Line data Source code
1 : /******************************************************************************************************
2 :
3 : (C) 2022-2026 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 <stdint.h>
34 : #include "options.h"
35 : #ifdef DEBUGGING
36 : #include "debug.h"
37 : #endif
38 : #include "cnst.h"
39 : #include "ivas_cnst.h"
40 : #include "ivas_prot.h"
41 : #include "rom_enc.h"
42 : #include "rom_com.h"
43 : #include "prot.h"
44 : #include "wmc_auto.h"
45 :
46 :
47 : /*-------------------------------------------------------------------*
48 : * pre_proc_ivas()
49 : *
50 : * Pre-processing (Selection of internal Fs, classification, SC VBR decision,
51 : * Decision matrix, Preprocessing at other Fs, core switching decision, ...)
52 : *--------------------------------------------------------------------*/
53 :
54 17835655 : void pre_proc_ivas(
55 : Encoder_State *st, /* i/o: encoder state structure */
56 : const int16_t last_element_mode, /* i : last element mode */
57 : const int32_t element_brate, /* i : element bitrate */
58 : const int32_t last_element_brate, /* i : last element bitrate */
59 : const int16_t input_frame, /* i : frame length */
60 : float old_inp_12k8[], /* i/o: buffer of old input signal */
61 : float old_inp_16k[], /* i/o: buffer of old input signal @ 16kHz */
62 : float **inp, /* o : ptr. to inp. signal in the current frame*/
63 : float *ener, /* o : residual energy from Levinson-Durbin */
64 : float A[NB_SUBFR16k * ( M + 1 )], /* i/o: A(z) unquantized for the 4 subframes */
65 : float Aw[NB_SUBFR16k * ( M + 1 )], /* i/o: weighted A(z) unquantized for subframes */
66 : float epsP[M + 1], /* i/o: LP prediction errors */
67 : float lsp_new[M], /* i/o: LSPs at the end of the frame */
68 : float lsp_mid[M], /* i/o: LSPs in the middle of the frame */
69 : float *new_inp_resamp16k, /* o : new input signal @16kHz, non pre-emphasised, used by the WB TBE/BWE */
70 : int16_t *Voicing_flag, /* o : voicing flag for HQ FEC */
71 : const float old_wsp[], /* i : weighted input signal buffer */
72 : const int16_t loc_harm, /* i : harmonicity flag */
73 : const float cor_map_sum, /* i : speech/music clasif. parameter */
74 : const int16_t vad_flag_dtx, /* i : HE-SAD flag with additional DTX HO */
75 : const float enerBuffer[CLDFB_NO_CHANNELS_MAX], /* i : energy buffer */
76 : const float fft_buff[2 * L_FFT], /* i : FFT buffer */
77 : const int16_t MCT_flag, /* i : hMCT handle allocated (1) or not (0) */
78 : const int16_t vad_hover_flag, /* i : VAD hangover flag */
79 : const int16_t flag_16k_smc /* i : flag to indicate if the OL SMC is run at 16 kHz */
80 : )
81 : {
82 : int16_t L_look, element_mode, lMemRecalc_12k8;
83 : float *inp_12k8, *new_inp_12k8, *inp_16k; /* pointers to current frame and new data */
84 : const float *wsp; /* weighted input signal buffer */
85 : int32_t sr_core_tmp, total_brate_tmp;
86 :
87 17835655 : push_wmops( "pre_proc" );
88 :
89 : /*----------------------------------------------------------------*
90 : * Initialization
91 : *----------------------------------------------------------------*/
92 :
93 17835655 : element_mode = st->element_mode;
94 :
95 17835655 : new_inp_12k8 = old_inp_12k8 + L_INP_MEM; /* pointer to new samples of the input signal in 12.8kHz core */
96 17835655 : inp_12k8 = new_inp_12k8 - L_LOOK_12k8;
97 :
98 17835655 : if ( element_mode != IVAS_CPE_DFT )
99 : {
100 16922866 : new_inp_12k8 -= L_FILT;
101 : }
102 :
103 17835655 : wsp = old_wsp + L_WSP_MEM; /* pointer to the current frame of weighted signal in 12.8kHz core */
104 :
105 17835655 : lMemRecalc_12k8 = 0;
106 17835655 : if ( element_mode == IVAS_CPE_TD )
107 : {
108 62354 : lMemRecalc_12k8 = NS2SA( INT_FS_12k8, L_MEM_RECALC_NS );
109 : }
110 :
111 : /*----------------------------------------------------------------*
112 : * Selection of internal ACELP Fs (12.8 kHz or 16 kHz)
113 : *----------------------------------------------------------------*/
114 :
115 17835655 : if ( st->core_brate == FRAME_NO_DATA )
116 : {
117 : /* prevent "L_frame" changes in CNG segments */
118 84741 : st->L_frame = st->last_L_frame;
119 : }
120 17750914 : else if ( st->core_brate == SID_2k40 && st->bwidth >= WB && st->hDtxEnc->first_CNG && ( st->hTdCngEnc != NULL && st->hTdCngEnc->act_cnt2 < MIN_ACT_CNG_UPD ) )
121 : {
122 : /* prevent "L_frame" changes in SID frame after short segment of active frames */
123 5820 : st->L_frame = st->hDtxEnc->last_CNG_L_frame;
124 : }
125 17745094 : else if ( ( ( st->element_mode == IVAS_CPE_MDCT && st->element_brate >= IVAS_64k && st->bwidth >= SWB ) || ( element_mode == IVAS_SCE && st->total_brate > MAX_ACELP_BRATE && st->bwidth >= SWB ) ) && st->core_brate != SID_2k40 )
126 : {
127 10157409 : st->L_frame = L_FRAME32k;
128 : }
129 7587685 : else if ( st->bwidth >= SWB && st->total_brate > MAX_ACELP_BRATE_ISM && st->total_brate <= MAX_ACELP_BRATE && element_mode == IVAS_SCE && st->is_ism_format && st->tcxonly && st->core_brate != SID_2k40 )
130 : {
131 56430 : st->L_frame = L_FRAME25_6k;
132 : }
133 7531255 : else if ( st->flag_ACELP16k )
134 : {
135 5768418 : st->L_frame = L_FRAME16k;
136 : }
137 : else
138 : {
139 1762837 : st->L_frame = L_FRAME;
140 : }
141 :
142 17835655 : if ( st->hFdCngEnc != NULL && st->element_mode != IVAS_CPE_MDCT && ( ( st->hFdCngEnc->hFdCngCom->frameSize != st->L_frame ) || ( st->hFdCngEnc->hFdCngCom->CngBandwidth != st->input_bwidth ) ) )
143 : {
144 413322 : configureFdCngEnc( st->hFdCngEnc, max( st->input_bwidth, WB ), st->L_frame == L_FRAME16k ? ACELP_16k40 : ACELP_9k60 );
145 : }
146 :
147 17835655 : if ( st->ini_frame == 0 )
148 : {
149 : /* avoid switching of internal ACELP Fs in the very first frame */
150 154587 : st->last_L_frame = st->L_frame;
151 : }
152 :
153 17835655 : if ( st->L_frame == L_FRAME )
154 : {
155 1792495 : st->gamma = GAMMA1;
156 1792495 : st->preemph_fac = PREEMPH_FAC;
157 : }
158 16043160 : else if ( st->L_frame == L_FRAME32k )
159 : {
160 10157409 : st->gamma = GAMMA16k;
161 10157409 : st->preemph_fac = PREEMPH_FAC_SWB;
162 : }
163 : else
164 : {
165 5885751 : st->gamma = GAMMA16k;
166 5885751 : st->preemph_fac = PREEMPH_FAC_16k;
167 : }
168 :
169 17835655 : st->sr_core = st->L_frame * FRAMES_PER_SEC;
170 17835655 : st->encoderLookahead_enc = NS2SA( st->sr_core, ACELP_LOOK_NS );
171 17835655 : st->encoderPastSamples_enc = ( st->L_frame * 9 ) >> 4;
172 :
173 :
174 : /*-----------------------------------------------------------------*
175 : * coder_type rewriting in case of switching
176 : * IC frames selection
177 : * enforce TC frames in case of switching
178 : *-----------------------------------------------------------------*/
179 :
180 : /* enforce TRANSITION frames */
181 17835655 : if ( !( st->element_mode == IVAS_CPE_TD && st->idchan == 1 ) && st->last_L_frame != st->L_frame && st->core_brate != FRAME_NO_DATA && st->core_brate != SID_2k40 && st->last_core_brate != FRAME_NO_DATA && st->last_core_brate != SID_2k40 && st->coder_type_raw != VOICED )
182 : {
183 : /* enforce TC frame in case of ACELP@12k8 <-> ACELP@16k core switching */
184 4319618 : st->coder_type = TRANSITION;
185 : }
186 13516037 : else if ( st->last_core == HQ_CORE && st->coder_type_raw != VOICED )
187 : {
188 : /* enforce TC frame in case of HQ -> ACELP core switching */
189 59282 : st->coder_type = TRANSITION;
190 : }
191 13456755 : else if ( st->last_core_brate <= SID_2k40 && st->cng_type == FD_CNG && !( element_mode == IVAS_CPE_TD ) )
192 : {
193 : /* enforce TC frame in case of FD_CNG -> ACELP switching (past excitation not available) */
194 93110 : st->coder_type = TRANSITION;
195 : }
196 : /* select INACTIVE frames */
197 13363645 : else if ( st->total_brate <= MAX_GSC_INACTIVE_BRATE && st->vad_flag == 0 && st->element_mode != IVAS_CPE_MDCT )
198 : {
199 : /* inactive frames will be coded by GSC technology */
200 : /* except for the VBR mode. VBR mode uses NELP for that */
201 531767 : if ( !( st->Opt_SC_VBR && vad_flag_dtx ) && ( st->idchan == 0 || element_mode != IVAS_CPE_TD ) )
202 : {
203 525657 : st->coder_type = INACTIVE;
204 525657 : st->hGSCEnc->noise_lev = NOISE_LEVEL_SP3;
205 : }
206 : }
207 12831878 : else if ( st->total_brate > MAX_GSC_INACTIVE_BRATE && ( ( st->vad_flag == 0 && st->bwidth >= SWB && st->max_bwidth >= SWB ) || ( st->localVAD == 0 && ( st->bwidth <= WB || st->max_bwidth <= WB ) ) ) )
208 : {
209 : /* inactive frames will be coded by AVQ technology (exceptionally it can be later rewritten to GSC technology in ivas_combined_format_brate_sanity()) */
210 1631678 : st->coder_type = INACTIVE;
211 : }
212 :
213 : /*---------------------------------------------------------------------*
214 : * Decision matrix (selection of technologies)
215 : *---------------------------------------------------------------------*/
216 :
217 17835655 : st->mdct_sw = MODE1;
218 17835655 : st->mdct_sw_enable = MODE1;
219 17835655 : if ( ( st->total_brate <= MIN_BRATE_GSC_NOISY_FLAG || st->bwidth < SWB || st->flag_ACELP16k ) && st->GSC_IVAS_mode == 0 )
220 : {
221 16909528 : st->GSC_noisy_speech = 0;
222 : }
223 :
224 : /* core selection */
225 17835655 : ivas_decision_matrix_enc( st, element_brate, fft_buff, enerBuffer, last_element_mode );
226 :
227 17835655 : if ( st->L_frame == L_FRAME16k && ( st->coder_type == VOICED || st->coder_type == UNVOICED ) ) /* VOICED and UNVOICED are not supported in ACELP@16k */
228 : {
229 0 : st->coder_type = GENERIC;
230 : }
231 :
232 17835655 : if ( st->core == TCX_20_CORE || st->core == HQ_CORE )
233 : {
234 15880524 : st->Nb_ACELP_frames = 0;
235 : /* Configure TCX with the same bitrate as given when (re-)initializing TCX */
236 15880524 : total_brate_tmp = st->total_brate;
237 15880524 : st->total_brate = st->bits_frame_nominal * FRAMES_PER_SEC;
238 15880524 : SetModeIndex( st, st->last_bits_frame_nominal * FRAMES_PER_SEC, last_element_mode, MCT_flag );
239 :
240 15880524 : st->sr_core = getCoreSamplerateMode2( element_mode, st->total_brate, st->bwidth, st->flag_ACELP16k, st->rf_mode, st->is_ism_format );
241 15880524 : st->total_brate = total_brate_tmp;
242 :
243 15880524 : st->L_frame = (int16_t) ( st->sr_core / FRAMES_PER_SEC );
244 15880524 : st->encoderLookahead_enc = NS2SA( st->sr_core, ACELP_LOOK_NS );
245 15880524 : st->encoderPastSamples_enc = ( st->L_frame * 9 ) >> 4;
246 :
247 15880524 : if ( st->sr_core == INT_FS_12k8 )
248 : {
249 706857 : st->preemph_fac = PREEMPH_FAC;
250 706857 : st->gamma = GAMMA1;
251 : }
252 15173667 : else if ( st->sr_core == INT_FS_16k )
253 : {
254 3035853 : st->preemph_fac = PREEMPH_FAC_16k;
255 3035853 : st->gamma = GAMMA16k;
256 : }
257 : else /* st->sr_core >=25600 */
258 : {
259 12137814 : st->preemph_fac = PREEMPH_FAC_SWB;
260 12137814 : st->gamma = GAMMA16k;
261 : }
262 :
263 15880524 : if ( st->vad_flag == 0 )
264 : {
265 2688773 : st->coder_type = INACTIVE;
266 : }
267 13191751 : else if ( st->coder_type > GENERIC )
268 : {
269 3859268 : st->coder_type = GENERIC;
270 : }
271 15880524 : if ( st->element_mode != IVAS_CPE_MDCT )
272 : {
273 3142934 : SetTCXModeInfo( st, st->hTranDet, &st->hTcxCfg->tcx_curr_overlap_mode );
274 : }
275 : }
276 1955131 : else if ( st->element_mode == IVAS_CPE_MDCT )
277 : {
278 31156 : st->hTcxEnc->tfm_mem = 0.75f;
279 : }
280 1923975 : else if ( element_brate != last_element_brate )
281 : {
282 626 : if ( st->core_brate != FRAME_NO_DATA )
283 : {
284 619 : SetModeIndex( st, st->bits_frame_nominal * FRAMES_PER_SEC, element_mode, MCT_flag );
285 : }
286 :
287 626 : if ( st->extl != -1 && st->extl != IGF_BWE && st->igf == 1 )
288 : {
289 619 : st->igf = 0;
290 : }
291 : }
292 :
293 : /*-----------------------------------------------------------------*
294 : * Update of ACELP harmonicity counter (used in ACELP transform codebook @32kbps)
295 : *-----------------------------------------------------------------*/
296 :
297 17835655 : if ( loc_harm == 1 && cor_map_sum > 50 && st->clas == VOICED_CLAS && st->coder_type == GENERIC )
298 : {
299 2843007 : st->last_harm_flag_acelp++;
300 :
301 2843007 : if ( st->last_harm_flag_acelp > 10 )
302 : {
303 753791 : st->last_harm_flag_acelp = 10;
304 : }
305 : }
306 : else
307 : {
308 14992648 : st->last_harm_flag_acelp = 0;
309 : }
310 :
311 : /*-----------------------------------------------------------------*
312 : * Update audio frames counter (used for UV decision)
313 : *-----------------------------------------------------------------*/
314 :
315 17835655 : if ( st->coder_type == AUDIO )
316 : {
317 91470 : st->audio_frame_cnt += AUDIO_COUNTER_STEP;
318 : }
319 17744185 : else if ( st->coder_type != INACTIVE )
320 : {
321 14681670 : st->audio_frame_cnt--;
322 : }
323 :
324 17835655 : if ( st->audio_frame_cnt > AUDIO_COUNTER_MAX )
325 : {
326 23411 : st->audio_frame_cnt = AUDIO_COUNTER_MAX;
327 : }
328 :
329 17835655 : if ( st->audio_frame_cnt < 0 )
330 : {
331 10735724 : st->audio_frame_cnt = 0;
332 : }
333 :
334 : /*-----------------------------------------------------------------*
335 : * Set formant sharpening flag
336 : *-----------------------------------------------------------------*/
337 :
338 17835655 : st->sharpFlag = 0;
339 :
340 17835655 : if ( st->coder_type == GENERIC || st->coder_type == VOICED || st->coder_type == TRANSITION )
341 : {
342 14634408 : if ( element_brate >= FRMT_SHP_MIN_BRATE_IVAS && st->lp_noise > FORMANT_SHARPENING_NOISE_THRESHOLD )
343 : {
344 1079048 : st->sharpFlag = 0;
345 : }
346 : else
347 : {
348 13555360 : st->sharpFlag = 1;
349 : }
350 : }
351 :
352 : /* TD stereo, secondary channel - due to lack of signaling bits, sharpFlag is always 1 */
353 17835655 : if ( element_mode == IVAS_CPE_TD && st->idchan == 1 )
354 : {
355 31177 : st->sharpFlag = 0;
356 31177 : if ( st->coder_type == GENERIC || st->coder_type == VOICED )
357 : {
358 23825 : st->sharpFlag = 1;
359 : }
360 : }
361 :
362 : /*-----------------------------------------------------------------*
363 : * Set voicing flag for HQ FEC
364 : *-----------------------------------------------------------------*/
365 :
366 17835655 : if ( st->sp_aud_decision1 == 0 && ( st->coder_type == VOICED || st->coder_type == GENERIC ) )
367 : {
368 7233190 : *Voicing_flag = 1;
369 : }
370 : else
371 : {
372 10602465 : *Voicing_flag = 0;
373 : }
374 :
375 : /*-----------------------------------------------------------------*
376 : * Compute core-coder buffers at internal sampling rate
377 : *-----------------------------------------------------------------*/
378 :
379 17835655 : sr_core_tmp = ( st->tcxonly == 0 ) ? INT_FS_16k : max( INT_FS_16k, st->sr_core ); /* indicates the ACELP sampling rate */
380 :
381 17835655 : L_look = NS2SA( sr_core_tmp, ACELP_LOOK_NS ); /* lookahead at other sampling rate (16kHz, 25.6kHz, 32kHz) */
382 :
383 17835655 : inp_16k = old_inp_16k + L_INP_MEM - L_look;
384 :
385 17835655 : if ( !flag_16k_smc )
386 : {
387 17212482 : ivas_compute_core_buffers( st, &inp_16k, old_inp_16k, new_inp_resamp16k, input_frame, last_element_mode, sr_core_tmp, ener, A, Aw, epsP, lsp_new, lsp_mid );
388 : }
389 :
390 17835655 : if ( !( st->L_frame == L_FRAME16k && element_mode != IVAS_CPE_MDCT ) )
391 : {
392 : /* update signal buffers */
393 15544430 : mvr2r( new_inp_12k8, st->buf_speech_enc_pe + st->L_frame, L_FRAME );
394 15544430 : mvr2r( st->buf_speech_enc + L_FRAME32k, st->buf_speech_enc + st->L_frame, L_FRAME );
395 :
396 15544430 : if ( element_mode == IVAS_CPE_DFT )
397 : {
398 421782 : mvr2r( st->buf_speech_enc + L_FRAME32k - STEREO_DFT_OVL_12k8, st->buf_speech_enc + st->L_frame - STEREO_DFT_OVL_12k8, STEREO_DFT_OVL_12k8 );
399 : }
400 15122648 : else if ( element_mode == IVAS_CPE_TD || element_mode == IVAS_CPE_MDCT )
401 : {
402 12808118 : mvr2r( st->buf_speech_enc + L_FRAME32k - lMemRecalc_12k8 - L_FILT, st->buf_speech_enc + st->L_frame - lMemRecalc_12k8 - L_FILT, lMemRecalc_12k8 + L_FILT );
403 : }
404 2314530 : else if ( element_mode == IVAS_SCE )
405 : {
406 2314530 : mvr2r( st->buf_speech_enc + L_FRAME32k - L_FILT, st->buf_speech_enc + st->L_frame - L_FILT, L_FILT );
407 : }
408 :
409 15544430 : if ( st->tcxonly == 0 )
410 : {
411 1722161 : mvr2r( wsp, st->wspeech_enc, L_FRAME + L_LOOK_12k8 );
412 : }
413 : }
414 :
415 17835655 : if ( flag_16k_smc )
416 : {
417 623173 : mvr2r( st->buf_speech_enc + L_FRAME16k, new_inp_resamp16k, L_FRAME16k );
418 : }
419 :
420 : /*-----------------------------------------------------------------*
421 : * Updates
422 : *-----------------------------------------------------------------*/
423 :
424 : /* update old weighted speech buffer - for OL pitch analysis */
425 17835655 : mvr2r( &old_wsp[L_FRAME], st->old_wsp, L_WSP_MEM );
426 :
427 : /* set the pointer of the current frame for the ACELP core */
428 17835655 : if ( st->L_frame == L_FRAME )
429 : {
430 1792495 : *inp = inp_12k8;
431 : }
432 : else
433 : {
434 16043160 : *inp = inp_16k;
435 : }
436 :
437 : /* Update VAD hangover frame counter in active frames */
438 17835655 : if ( !( st->core_brate == SID_2k40 || st->core_brate == FRAME_NO_DATA ) && st->tcxonly == 0 )
439 : {
440 3905603 : if ( st->hTdCngEnc != NULL && st->Opt_DTX_ON && vad_hover_flag )
441 : {
442 14011 : st->hTdCngEnc->burst_ho_cnt++;
443 14011 : if ( st->hTdCngEnc->burst_ho_cnt > HO_HIST_SIZE )
444 : {
445 1278 : st->hTdCngEnc->burst_ho_cnt = HO_HIST_SIZE;
446 : }
447 : }
448 3891592 : else if ( st->hTdCngEnc != NULL && vad_flag_dtx )
449 : {
450 670705 : st->hTdCngEnc->burst_ho_cnt = 0;
451 : }
452 : }
453 :
454 : #ifdef DEBUG_MODE_ACELP
455 : dbgwrite( inp_12k8, sizeof( float ), L_FRAME, 1, fname( debug_dir, "inp_12k8", st->idchan, st->id_element, ENC ) );
456 : dbgwrite( inp_16k, sizeof( float ), L_FRAME, 1, fname( debug_dir, "inp_16k", st->idchan, st->id_element, ENC ) );
457 : #endif
458 :
459 17835655 : pop_wmops();
460 17835655 : return;
461 : }
462 :
463 :
464 : /*-------------------------------------------------------------------*
465 : * ivas_compute_core_buffers()
466 : *
467 : * Compute core-coder buffers at internal sampling rate
468 : *--------------------------------------------------------------------*/
469 :
470 17835655 : void ivas_compute_core_buffers(
471 : Encoder_State *st, /* i/o: encoder state structure */
472 : float **inp16k_out, /* o : ptr. to inp. signal in the current frame */
473 : float *old_inp_16k, /* i/o: buffer of old input signal @ 16kHz */
474 : float new_inp_resamp16k_out[], /* o : new input signal @16kHz, non pre-emphasised, used by the WB TBE/BWE */
475 : const int16_t input_frame, /* i : frame length */
476 : const int16_t last_element_mode, /* i : last element mode */
477 : const int32_t sr_core, /* i : core-coder sampling rate */
478 : float *ener, /* o : residual energy from Levinson-Durbin */
479 : float A[NB_SUBFR16k * ( M + 1 )], /* i/o: A(z) unquantized for the 4 subframes */
480 : float Aw[NB_SUBFR16k * ( M + 1 )], /* i/o: weighted A(z) unquantized for subframes */
481 : float epsP[M + 1], /* i/o: LP prediction errors */
482 : float lsp_new[M], /* i/o: LSPs at the end of the frame */
483 : float lsp_mid[M] /* i/o: LSPs in the middle of the frame */
484 : )
485 : {
486 : float *inp_16k, *new_inp_16k;
487 : float tmp, mem_decim16k_dummy[2 * L_FILT_MAX];
488 : const float *signal_in;
489 : int16_t delay, element_mode;
490 : float temp1F_icatdmResampBuf[L_FILT_MAX]; /* temp buffers for ICA TDM resamplers */
491 : float new_inp_resamp16k[L_FRAME16k];
492 : int16_t lMemRecalc, lMemRecalc_16k, L_frame_tmp, L_look;
493 : int32_t input_Fs;
494 :
495 17835655 : signal_in = st->input;
496 :
497 17835655 : input_Fs = st->input_Fs;
498 17835655 : element_mode = st->element_mode;
499 :
500 17835655 : lMemRecalc_16k = 0;
501 17835655 : lMemRecalc = 0;
502 17835655 : if ( element_mode == IVAS_CPE_TD )
503 : {
504 62354 : lMemRecalc_16k = NS2SA( INT_FS_16k, L_MEM_RECALC_NS );
505 62354 : lMemRecalc = NS2SA( input_Fs, L_MEM_RECALC_NS );
506 : }
507 :
508 : /*---------------------------------------------------------------*
509 : * Preprocessing at other sampling frequency rate (16/25.6/32kHz)
510 : *----------------------------------------------------------------*/
511 :
512 17835655 : L_frame_tmp = ( st->tcxonly == 0 ) ? L_FRAME16k : max( L_FRAME16k, st->L_frame );
513 :
514 17835655 : L_look = NS2SA( sr_core, ACELP_LOOK_NS ); /* lookahead at other sampling rate (16kHz, 25.6kHz, 32kHz) */
515 :
516 17835655 : new_inp_16k = old_inp_16k + L_INP_MEM; /* pointer to new samples of the input signal in 16kHz core */
517 17835655 : inp_16k = new_inp_16k - L_look; /* pointer to the current frame of input signal in 16kHz core */
518 :
519 : /* shift the pointer back to take care of resampler memory update */
520 17835655 : if ( element_mode == IVAS_CPE_TD || element_mode == IVAS_SCE )
521 : {
522 4154120 : new_inp_16k -= NS2SA( sr_core, DELAY_FIR_RESAMPL_NS );
523 : }
524 :
525 17835655 : if ( element_mode == IVAS_CPE_DFT )
526 : {
527 912789 : mvr2r( st->old_inp_16k, old_inp_16k, L_INP_MEM - STEREO_DFT_OVL_16k );
528 : }
529 16922866 : else if ( element_mode == IVAS_CPE_TD )
530 : {
531 62354 : mvr2r( st->old_inp_16k, old_inp_16k, L_INP_MEM - L_MEM_RECALC_16K - L_FILT16k );
532 : }
533 : else
534 : {
535 16860512 : mvr2r( st->old_inp_16k, old_inp_16k, L_INP_MEM - L_FILT16k );
536 : }
537 :
538 : /*---------------------------------------------------------------*
539 : * Change the sampling frequency to 16/25.6/32 kHz
540 : *----------------------------------------------------------------*/
541 :
542 17835655 : if ( element_mode == IVAS_SCE )
543 : {
544 4091766 : if ( input_Fs == sr_core )
545 : {
546 : /* no resampling needed, only delay adjustment to account for the FIR resampling delay */
547 230120 : delay = NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS );
548 230120 : mvr2r( st->mem_decim16k + delay, new_inp_16k, delay );
549 230120 : mvr2r( signal_in, new_inp_16k + delay, input_frame );
550 230120 : mvr2r( signal_in + input_frame - 2 * delay, st->mem_decim16k, 2 * delay );
551 : }
552 3861646 : else if ( input_Fs == 32000 || input_Fs == 48000 )
553 : {
554 3861646 : modify_Fs( signal_in, input_frame, input_Fs, new_inp_16k, sr_core, st->mem_decim16k, 0 );
555 :
556 3861646 : mvr2r( st->mem_decim16k, mem_decim16k_dummy, 2 * L_FILT_MAX );
557 3861646 : set_f( temp1F_icatdmResampBuf, 0, L_FILT_MAX );
558 3861646 : modify_Fs( temp1F_icatdmResampBuf, NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS ), input_Fs, new_inp_16k + NS2SA( sr_core, FRAME_SIZE_NS ), sr_core, mem_decim16k_dummy, 0 );
559 : }
560 : }
561 13743889 : else if ( element_mode == IVAS_CPE_TD )
562 : {
563 62354 : if ( input_Fs == sr_core )
564 : {
565 : /* no resampling needed, only delay adjustment to account for the FIR resampling delay */
566 5294 : delay = NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS );
567 5294 : mvr2r( st->mem_decim16k + delay, new_inp_16k - lMemRecalc - delay + L_FILT16k, delay );
568 5294 : mvr2r( signal_in - lMemRecalc, new_inp_16k - lMemRecalc + L_FILT16k, input_frame + lMemRecalc );
569 5294 : mvr2r( signal_in + input_frame - lMemRecalc - 2 * delay, st->mem_decim16k, 2 * delay );
570 : }
571 57060 : else if ( input_Fs == 32000 || input_Fs == 48000 )
572 : {
573 : /* reconstruct past segment of input signal when switching from MDCT stereo */
574 57060 : if ( last_element_mode == IVAS_CPE_MDCT /*|| st->idchan == 1*/ )
575 : {
576 484 : int16_t length_inp = NS2SA( input_Fs, L_MEM_RECALC_SCH_NS - DELAY_FIR_RESAMPL_NS );
577 484 : int16_t length_16k = NS2SA( INT_FS_16k, L_MEM_RECALC_SCH_NS - DELAY_FIR_RESAMPL_NS );
578 :
579 484 : mvr2r( signal_in - lMemRecalc - length_inp - 2 * NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS ), st->mem_decim16k, 2 * NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS ) );
580 484 : modify_Fs( signal_in - lMemRecalc - length_inp, length_inp, input_Fs, new_inp_16k - lMemRecalc_16k - length_16k, sr_core, st->mem_decim16k, 0 );
581 : }
582 :
583 57060 : modify_Fs( signal_in - lMemRecalc, input_frame, input_Fs, new_inp_16k - lMemRecalc_16k, sr_core, st->mem_decim16k, 0 );
584 57060 : mvr2r( st->mem_decim16k, mem_decim16k_dummy, 2 * L_FILT_MAX );
585 :
586 57060 : if ( lMemRecalc > 0 )
587 : {
588 57060 : modify_Fs( signal_in - lMemRecalc + input_frame, lMemRecalc, input_Fs, new_inp_16k + NS2SA( sr_core, FRAME_SIZE_NS ) - lMemRecalc_16k, sr_core, mem_decim16k_dummy, 0 );
589 : }
590 :
591 57060 : set_f( temp1F_icatdmResampBuf, 0, L_FILT_MAX );
592 57060 : modify_Fs( temp1F_icatdmResampBuf, NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS ), input_Fs, new_inp_16k + NS2SA( sr_core, FRAME_SIZE_NS ), sr_core, mem_decim16k_dummy, 0 );
593 : }
594 : }
595 13681535 : else if ( st->idchan == 0 )
596 : {
597 : /* update the FIR resampling filter memory, needed for switching to time-domain (FIR) resampling */
598 7297162 : mvr2r( signal_in + input_frame - NS2SA( input_Fs, L_MEM_RECALC_NS ) - 2 * NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS ), st->mem_decim16k, 2 * NS2SA( input_Fs, DELAY_FIR_RESAMPL_NS ) );
599 : }
600 :
601 : /*------------------------------------------------*
602 : * Update BWE memories *
603 : *------------------------------------------------*/
604 :
605 17835655 : if ( sr_core == INT_FS_16k && element_mode != IVAS_CPE_MDCT )
606 : {
607 4083697 : delay = NS2SA( INT_FS_16k, DELAY_FD_BWE_ENC_12k8_NS );
608 :
609 4083697 : if ( element_mode == IVAS_CPE_DFT )
610 : {
611 : /* save input resampled at 16kHz, non-preemphasised.*/
612 912789 : mvr2r( new_inp_16k, new_inp_resamp16k, L_FRAME16k );
613 :
614 912789 : if ( st->bwidth == WB )
615 : {
616 189063 : mvr2r( new_inp_16k - delay, st->hBWE_FD->old_input_wb, delay );
617 189063 : mvr2r( new_inp_16k - STEREO_DFT_OVL_16k, st->hBWE_FD->old_wtda_swb + L_FRAME16k - STEREO_DFT_OVL_16k + delay, STEREO_DFT_OVL_16k - delay );
618 : }
619 : }
620 3170908 : else if ( element_mode == IVAS_CPE_TD )
621 : {
622 : /* save input resampled at 16kHz, non-preemphasised */
623 62354 : mvr2r( new_inp_16k + L_FILT16k, new_inp_resamp16k, L_FRAME16k );
624 :
625 62354 : if ( st->bwidth == WB && st->hBWE_FD != NULL )
626 : {
627 5819 : mvr2r( new_inp_16k + L_FILT16k - delay, st->hBWE_FD->old_input_wb, delay );
628 5819 : mvr2r( new_inp_16k - L_MEM_RECALC_16K, st->hBWE_FD->old_wtda_swb + L_FRAME16k - L_MEM_RECALC_16K - L_FILT16k + delay, L_MEM_RECALC_16K + L_FILT16k - delay );
629 : }
630 : }
631 3108554 : else if ( element_mode == IVAS_SCE )
632 : {
633 : /* save input resampled at 16kHz, non-preemphasised */
634 3108554 : mvr2r( new_inp_16k + L_FILT16k, new_inp_resamp16k, L_FRAME16k );
635 :
636 3108554 : if ( st->bwidth == WB )
637 : {
638 630566 : mvr2r( new_inp_16k, st->hBWE_FD->old_input_wb + delay - L_FILT16k, L_FILT16k );
639 : /* all buffer st->hBWE_FD->old_wtda_swb is correct and does not need to be updated */
640 : }
641 : }
642 : }
643 13751958 : else if ( sr_core > INT_FS_16k )
644 : {
645 : /* reset the buffer, the signal is needed for WB BWEs */
646 12137814 : set_f( new_inp_resamp16k, 0.0f, L_FRAME16k );
647 : }
648 :
649 : /*------------------------------------------------------------------*
650 : * Perform fixed preemphasis (16kHz signal) through 1 - g*z^-1
651 : *-----------------------------------------------------------------*/
652 :
653 17835655 : if ( st->tcxonly == 0 && !( ( element_mode == IVAS_CPE_TD && st->idchan == 1 ) || element_mode == IVAS_CPE_MDCT ) )
654 : {
655 3942909 : st->mem_preemph_enc = new_inp_16k[L_frame_tmp - 1];
656 : }
657 :
658 17835655 : if ( sr_core == INT_FS_16k && element_mode != IVAS_CPE_MDCT )
659 : {
660 4083697 : if ( element_mode == IVAS_CPE_DFT )
661 : {
662 912789 : mvr2r( new_inp_16k - STEREO_DFT_OVL_16k + L_FRAME16k, st->inp_16k_mem_stereo_sw, STEREO_DFT_OVL_16k - L_MEM_RECALC_16K - L_FILT16k ); /* update for TD/DFT stereo switching */
663 :
664 912789 : st->mem_preemph16k = st->mem_preemph16k_DFT;
665 912789 : st->mem_preemph16k_DFT = old_inp_16k[L_INP_MEM - STEREO_DFT_OVL_16k + L_FRAME16k - 1];
666 :
667 912789 : if ( st->L_frame == L_FRAME16k )
668 : {
669 491007 : mvr2r( new_inp_16k - STEREO_DFT_OVL_16k, st->buf_speech_enc + L_FRAME16k - STEREO_DFT_OVL_16k, L_FRAME16k + STEREO_DFT_OVL_16k );
670 : }
671 912789 : preemph( new_inp_16k - STEREO_DFT_OVL_16k, PREEMPH_FAC_16k, L_FRAME16k, &( st->mem_preemph16k ) );
672 912789 : tmp = st->mem_preemph16k;
673 912789 : preemph( new_inp_16k - STEREO_DFT_OVL_16k + L_FRAME16k, PREEMPH_FAC_16k, STEREO_DFT_OVL_16k, &tmp );
674 : }
675 3170908 : else if ( st->element_mode == IVAS_CPE_TD )
676 : {
677 62354 : if ( last_element_mode == IVAS_CPE_DFT )
678 : {
679 902 : st->mem_preemph16k = st->mem_preemph16k_DFT;
680 902 : mvr2r( st->inp_16k_mem_stereo_sw, new_inp_16k - L_MEM_RECALC_16K - ( STEREO_DFT_OVL_16k - L_MEM_RECALC_16K - L_FILT16k ), STEREO_DFT_OVL_16k - L_MEM_RECALC_16K - L_FILT16k );
681 902 : preemph( new_inp_16k - L_MEM_RECALC_16K - ( STEREO_DFT_OVL_16k - L_MEM_RECALC_16K - L_FILT16k ), PREEMPH_FAC_16k, STEREO_DFT_OVL_16k - L_MEM_RECALC_16K - L_FILT16k, &st->mem_preemph16k );
682 : }
683 :
684 62354 : st->mem_preemph16k_DFT = old_inp_16k[L_INP_MEM - STEREO_DFT_OVL_16k + L_FRAME16k - 1];
685 :
686 : /* preemphasise past segment of input signal when switching from MDCT stereo */
687 62354 : if ( last_element_mode == IVAS_CPE_MDCT )
688 : {
689 484 : int16_t length_16k = NS2SA( INT_FS_16k, L_MEM_RECALC_SCH_NS - DELAY_FIR_RESAMPL_NS );
690 484 : preemph( new_inp_16k - lMemRecalc_16k - length_16k, PREEMPH_FAC, length_16k, &st->mem_preemph16k );
691 : }
692 :
693 62354 : if ( st->L_frame == L_FRAME16k )
694 : {
695 22982 : mvr2r( new_inp_16k - lMemRecalc_16k, st->buf_speech_enc + L_FRAME16k - lMemRecalc_16k - L_FILT16k, L_FRAME16k + lMemRecalc_16k + L_FILT16k );
696 : }
697 62354 : preemph( new_inp_16k - lMemRecalc_16k, PREEMPH_FAC_16k, L_FRAME16k, &( st->mem_preemph16k ) );
698 62354 : tmp = st->mem_preemph16k;
699 62354 : preemph( new_inp_16k - lMemRecalc_16k + L_FRAME16k, PREEMPH_FAC_16k, lMemRecalc_16k + L_FILT16k, &tmp );
700 : }
701 3108554 : else if ( element_mode == IVAS_SCE )
702 : {
703 3108554 : preemph( new_inp_16k, PREEMPH_FAC_16k, L_FRAME16k, &( st->mem_preemph16k ) );
704 3108554 : tmp = st->mem_preemph16k;
705 3108554 : preemph( new_inp_16k + L_FRAME16k, PREEMPH_FAC_16k, L_FILT16k, &tmp );
706 : }
707 : }
708 : else /* keep memory up-to-date in case of bitrate switching */
709 : {
710 13751958 : if ( element_mode == IVAS_CPE_MDCT )
711 : {
712 12768746 : st->mem_preemph16k = 0;
713 : }
714 : else /* SCE */
715 : {
716 983212 : st->mem_preemph16k = new_inp_16k[L_frame_tmp - 1];
717 : }
718 13751958 : st->mem_preemph16k_DFT = st->mem_preemph16k;
719 : }
720 :
721 : /*-----------------------------------------------------------------*
722 : * LP analysis at 16kHz if ACELP@16k core was selected
723 : * update buffers
724 : *-----------------------------------------------------------------*/
725 :
726 17835655 : if ( st->L_frame == L_FRAME16k && element_mode != IVAS_CPE_MDCT )
727 : {
728 : /* update signal buffers */
729 2291225 : if ( element_mode == IVAS_CPE_DFT )
730 : {
731 491007 : mvr2r( new_inp_16k - STEREO_DFT_OVL_16k, st->buf_speech_enc_pe + L_FRAME16k - STEREO_DFT_OVL_16k, L_FRAME16k + STEREO_DFT_OVL_16k );
732 : }
733 1800218 : else if ( element_mode == IVAS_CPE_TD )
734 : {
735 22982 : mvr2r( new_inp_16k - lMemRecalc_16k, st->buf_speech_enc_pe + L_FRAME16k - lMemRecalc_16k - L_FILT16k, L_FRAME16k + lMemRecalc_16k + L_FILT16k );
736 : }
737 : else
738 : {
739 1777236 : mvr2r( new_inp_resamp16k, st->buf_speech_enc + L_FRAME16k, L_FRAME16k );
740 1777236 : mvr2r( new_inp_16k, st->buf_speech_enc_pe + L_FRAME16k, L_FRAME16k );
741 : }
742 :
743 : /*--------------------------------------------------------------*
744 : * LPC analysis
745 : *---------------------------------------------------------------*/
746 :
747 2291225 : if ( st->last_L_frame == L_FRAME )
748 : {
749 : /* this is just an approximation, but it is sufficient */
750 18366 : mvr2r( st->lsp_old1, st->lspold_enc, M );
751 : }
752 :
753 2291225 : analy_lp( inp_16k, L_FRAME16k, L_look, ener, A, epsP, lsp_new, lsp_mid, st->lspold_enc, st->pitch, st->voicing, INT_FS_16k, 0 );
754 :
755 : /*--------------------------------------------------------------*
756 : * Compute Weighted Input
757 : *---------------------------------------------------------------*/
758 :
759 2291225 : find_wsp( L_FRAME16k, L_SUBFR, NB_SUBFR16k, A, Aw, st->speech_enc_pe, PREEMPH_FAC_16k, st->wspeech_enc, &st->mem_wsp_enc, st->gamma, L_LOOK_16k );
760 : }
761 :
762 : /*-----------------------------------------------------------------*
763 : * Updates
764 : *-----------------------------------------------------------------*/
765 :
766 : /* update old input signal @16kHz buffer */
767 17835655 : if ( ( element_mode == IVAS_CPE_TD && st->idchan == 1 ) || element_mode == IVAS_CPE_MDCT )
768 : {
769 12799923 : set_f( st->old_inp_16k, 0, L_INP_MEM );
770 : }
771 5035732 : else if ( sr_core == INT_FS_16k )
772 : {
773 4052520 : mvr2r( &old_inp_16k[L_frame_tmp], st->old_inp_16k, L_INP_MEM );
774 : }
775 : else
776 : {
777 983212 : lerp( st->old_inp_12k8 + L_INP_MEM - L_INP_MEM * 4 / 5, st->old_inp_16k, L_INP_MEM, L_INP_MEM * 4 / 5 );
778 : }
779 :
780 17835655 : if ( inp16k_out != NULL )
781 : {
782 17212482 : *inp16k_out = inp_16k;
783 : }
784 :
785 17835655 : if ( new_inp_resamp16k_out != NULL )
786 : {
787 17212482 : mvr2r( new_inp_resamp16k, new_inp_resamp16k_out, L_FRAME16k );
788 : }
789 :
790 17835655 : return;
791 : }
|
