1 /* 2 * Copyright (c) 2012 The WebRTC project authors. All Rights Reserved. 3 * 4 * Use of this source code is governed by a BSD-style license 5 * that can be found in the LICENSE file in the root of the source 6 * tree. An additional intellectual property rights grant can be found 7 * in the file PATENTS. All contributing project authors may 8 * be found in the AUTHORS file in the root of the source tree. 9 */ 10 11 /* 12 * encode.c 13 * 14 * This file contains definition of funtions for encoding. 15 * Decoding of upper-band, including 8-12 kHz, when the bandwidth is 16 * 0-12 kHz, and 8-16 kHz, when the bandwidth is 0-16 kHz. 17 * 18 */ 19 20 #include <stdlib.h> 21 #include <string.h> 22 #include <stdio.h> 23 24 #include "structs.h" 25 #include "codec.h" 26 #include "pitch_estimator.h" 27 #include "entropy_coding.h" 28 #include "arith_routines.h" 29 #include "pitch_gain_tables.h" 30 #include "pitch_lag_tables.h" 31 #include "spectrum_ar_model_tables.h" 32 #include "lpc_tables.h" 33 #include "lpc_analysis.h" 34 #include "bandwidth_estimator.h" 35 #include "lpc_shape_swb12_tables.h" 36 #include "lpc_shape_swb16_tables.h" 37 #include "lpc_gain_swb_tables.h" 38 39 40 #define UB_LOOKAHEAD 24 41 42 43 /* 44 Rate allocation tables of lower and upper-band bottleneck for 45 12kHz & 16kHz bandwidth. 46 47 12 kHz bandwidth 48 ----------------- 49 The overall bottleneck of the coder is between 38 kbps and 45 kbps. We have 50 considered 7 enteries, uniformly distributed in this interval, i.e. 38, 51 39.17, 40.33, 41.5, 42.67, 43.83 and 45. For every entery, the lower-band 52 and the upper-band bottlenecks are specified in 53 'kLowerBandBitRate12' and 'kUpperBandBitRate12' 54 tables, respectively. E.g. the overall rate of 41.5 kbps corresponts to a 55 bottleneck of 31 kbps for lower-band and 27 kbps for upper-band. Given an 56 overall bottleneck of the codec, we use linear interpolation to get 57 lower-band and upper-band bottlenecks. 58 59 16 kHz bandwidth 60 ----------------- 61 The overall bottleneck of the coder is between 50 kbps and 56 kbps. We have 62 considered 7 enteries, uniformly distributed in this interval, i.e. 50, 51.2, 63 52.4, 53.6, 54.8 and 56. For every entery, the lower-band and the upper-band 64 bottlenecks are specified in 'kLowerBandBitRate16' and 65 'kUpperBandBitRate16' tables, respectively. E.g. the overall rate 66 of 53.6 kbps corresponts to a bottleneck of 32 kbps for lower-band and 30 67 kbps for upper-band. Given an overall bottleneck of the codec, we use linear 68 interpolation to get lower-band and upper-band bottlenecks. 69 70 */ 71 72 /* 38 39.17 40.33 41.5 42.67 43.83 45 */ 73 static const WebRtc_Word16 kLowerBandBitRate12[7] = { 74 29000, 30000, 30000, 31000, 31000, 32000, 32000 }; 75 static const WebRtc_Word16 kUpperBandBitRate12[7] = { 76 25000, 25000, 27000, 27000, 29000, 29000, 32000 }; 77 78 /* 50 51.2 52.4 53.6 54.8 56 */ 79 static const WebRtc_Word16 kLowerBandBitRate16[6] = { 80 31000, 31000, 32000, 32000, 32000, 32000 }; 81 static const WebRtc_Word16 kUpperBandBitRate16[6] = { 82 28000, 29000, 29000, 30000, 31000, 32000 }; 83 84 /****************************************************************************** 85 * WebRtcIsac_RateAllocation() 86 * Internal function to perform a rate-allocation for upper and lower-band, 87 * given a total rate. 88 * 89 * Input: 90 * - inRateBitPerSec : a total bottleneck in bits/sec. 91 * 92 * Output: 93 * - rateLBBitPerSec : a bottleneck allocated to the lower-band 94 * in bits/sec. 95 * - rateUBBitPerSec : a bottleneck allocated to the upper-band 96 * in bits/sec. 97 * 98 * Return value : 0 if rate allocation has been successful. 99 * -1 if failed to allocate rates. 100 */ 101 102 WebRtc_Word16 WebRtcIsac_RateAllocation(WebRtc_Word32 inRateBitPerSec, 103 double* rateLBBitPerSec, 104 double* rateUBBitPerSec, 105 enum ISACBandwidth* bandwidthKHz) { 106 WebRtc_Word16 idx; 107 double idxD; 108 double idxErr; 109 if (inRateBitPerSec < 38000) { 110 /* If the given overall bottleneck is less than 38000 then 111 * then codec has to operate in wideband mode, i.e. 8 kHz 112 * bandwidth. */ 113 *rateLBBitPerSec = (WebRtc_Word16)((inRateBitPerSec > 32000) ? 114 32000 : inRateBitPerSec); 115 *rateUBBitPerSec = 0; 116 *bandwidthKHz = isac8kHz; 117 } else if ((inRateBitPerSec >= 38000) && (inRateBitPerSec < 50000)) { 118 /* At a bottleneck between 38 and 50 kbps the codec is operating 119 * at 12 kHz bandwidth. Using xxxBandBitRate12[] to calculates 120 * upper/lower bottleneck */ 121 122 /* Find the bottlenecks by linear interpolation, 123 * step is (45000 - 38000)/6.0 we use the inverse of it. */ 124 const double stepSizeInv = 8.5714286e-4; 125 idxD = (inRateBitPerSec - 38000) * stepSizeInv; 126 idx = (idxD >= 6) ? 6 : ((WebRtc_Word16)idxD); 127 idxErr = idxD - idx; 128 *rateLBBitPerSec = kLowerBandBitRate12[idx]; 129 *rateUBBitPerSec = kUpperBandBitRate12[idx]; 130 131 if (idx < 6) { 132 *rateLBBitPerSec += (WebRtc_Word16)( 133 idxErr * (kLowerBandBitRate12[idx + 1] - kLowerBandBitRate12[idx])); 134 *rateUBBitPerSec += (WebRtc_Word16)( 135 idxErr * (kUpperBandBitRate12[idx + 1] - kUpperBandBitRate12[idx])); 136 } 137 *bandwidthKHz = isac12kHz; 138 } else if ((inRateBitPerSec >= 50000) && (inRateBitPerSec <= 56000)) { 139 /* A bottleneck between 50 and 56 kbps corresponds to bandwidth 140 * of 16 kHz. Using xxxBandBitRate16[] to calculates 141 * upper/lower bottleneck. */ 142 143 /* Find the bottlenecks by linear interpolation 144 * step is (56000 - 50000)/5 we use the inverse of it. */ 145 const double stepSizeInv = 8.3333333e-4; 146 idxD = (inRateBitPerSec - 50000) * stepSizeInv; 147 idx = (idxD >= 5) ? 5 : ((WebRtc_Word16)idxD); 148 idxErr = idxD - idx; 149 *rateLBBitPerSec = kLowerBandBitRate16[idx]; 150 *rateUBBitPerSec = kUpperBandBitRate16[idx]; 151 152 if (idx < 5) { 153 *rateLBBitPerSec += (WebRtc_Word16)(idxErr * 154 (kLowerBandBitRate16[idx + 1] - 155 kLowerBandBitRate16[idx])); 156 157 *rateUBBitPerSec += (WebRtc_Word16)(idxErr * 158 (kUpperBandBitRate16[idx + 1] - 159 kUpperBandBitRate16[idx])); 160 } 161 *bandwidthKHz = isac16kHz; 162 } else { 163 /* Out-of-range botlteneck value. */ 164 return -1; 165 } 166 167 /* limit the values. */ 168 *rateLBBitPerSec = (*rateLBBitPerSec > 32000) ? 32000 : *rateLBBitPerSec; 169 *rateUBBitPerSec = (*rateUBBitPerSec > 32000) ? 32000 : *rateUBBitPerSec; 170 return 0; 171 } 172 173 174 void WebRtcIsac_ResetBitstream(Bitstr* bit_stream) { 175 bit_stream->W_upper = 0xFFFFFFFF; 176 bit_stream->stream_index = 0; 177 bit_stream->streamval = 0; 178 } 179 180 int WebRtcIsac_EncodeLb(float* in, ISACLBEncStruct* ISACencLB_obj, 181 WebRtc_Word16 codingMode, 182 WebRtc_Word16 bottleneckIndex) { 183 int stream_length = 0; 184 int err; 185 int k; 186 int iterCntr; 187 188 double lofilt_coef[(ORDERLO + 1)*SUBFRAMES]; 189 double hifilt_coef[(ORDERHI + 1)*SUBFRAMES]; 190 float LP[FRAMESAMPLES_HALF]; 191 float HP[FRAMESAMPLES_HALF]; 192 193 double LP_lookahead[FRAMESAMPLES_HALF]; 194 double HP_lookahead[FRAMESAMPLES_HALF]; 195 double LP_lookahead_pf[FRAMESAMPLES_HALF + QLOOKAHEAD]; 196 double LPw[FRAMESAMPLES_HALF]; 197 198 double HPw[FRAMESAMPLES_HALF]; 199 double LPw_pf[FRAMESAMPLES_HALF]; 200 WebRtc_Word16 fre[FRAMESAMPLES_HALF]; /* Q7 */ 201 WebRtc_Word16 fim[FRAMESAMPLES_HALF]; /* Q7 */ 202 203 double PitchLags[4]; 204 double PitchGains[4]; 205 WebRtc_Word16 PitchGains_Q12[4]; 206 WebRtc_Word16 AvgPitchGain_Q12; 207 208 int frame_mode; /* 0 for 30ms, 1 for 60ms */ 209 int status = 0; 210 int my_index; 211 transcode_obj transcodingParam; 212 double bytesLeftSpecCoding; 213 WebRtc_UWord16 payloadLimitBytes; 214 215 /* Copy new frame-length and bottleneck rate only for the first 10 ms data */ 216 if (ISACencLB_obj->buffer_index == 0) { 217 /* Set the framelength for the next packet. */ 218 ISACencLB_obj->current_framesamples = ISACencLB_obj->new_framelength; 219 } 220 /* 'frame_mode' is 0 (30 ms) or 1 (60 ms). */ 221 frame_mode = ISACencLB_obj->current_framesamples / MAX_FRAMESAMPLES; 222 223 /* buffer speech samples (by 10ms packet) until the frame-length */ 224 /* is reached (30 or 60 ms). */ 225 /*****************************************************************/ 226 227 /* fill the buffer with 10ms input data */ 228 for (k = 0; k < FRAMESAMPLES_10ms; k++) { 229 ISACencLB_obj->data_buffer_float[k + ISACencLB_obj->buffer_index] = in[k]; 230 } 231 232 /* If buffersize is not equal to current framesize then increase index 233 * and return. We do no encoding untill we have enough audio. */ 234 if (ISACencLB_obj->buffer_index + FRAMESAMPLES_10ms != FRAMESAMPLES) { 235 ISACencLB_obj->buffer_index += FRAMESAMPLES_10ms; 236 return 0; 237 } 238 /* If buffer reached the right size, reset index and continue with 239 * encoding the frame. */ 240 ISACencLB_obj->buffer_index = 0; 241 242 /* End of buffer function. */ 243 /**************************/ 244 245 /* Encoding */ 246 /************/ 247 248 if (frame_mode == 0 || ISACencLB_obj->frame_nb == 0) { 249 /* This is to avoid Linux warnings until we change 'int' to 'Word32' 250 * at all places. */ 251 int intVar; 252 /* reset bitstream */ 253 WebRtcIsac_ResetBitstream(&(ISACencLB_obj->bitstr_obj)); 254 255 if ((codingMode == 0) && (frame_mode == 0) && 256 (ISACencLB_obj->enforceFrameSize == 0)) { 257 ISACencLB_obj->new_framelength = WebRtcIsac_GetNewFrameLength( 258 ISACencLB_obj->bottleneck, ISACencLB_obj->current_framesamples); 259 } 260 261 ISACencLB_obj->s2nr = WebRtcIsac_GetSnr( 262 ISACencLB_obj->bottleneck, ISACencLB_obj->current_framesamples); 263 264 /* Encode frame length. */ 265 status = WebRtcIsac_EncodeFrameLen( 266 ISACencLB_obj->current_framesamples, &ISACencLB_obj->bitstr_obj); 267 if (status < 0) { 268 /* Wrong frame size. */ 269 return status; 270 } 271 /* Save framelength for multiple packets memory. */ 272 ISACencLB_obj->SaveEnc_obj.framelength = 273 ISACencLB_obj->current_framesamples; 274 275 /* To be used for Redundant Coding. */ 276 ISACencLB_obj->lastBWIdx = bottleneckIndex; 277 intVar = (int)bottleneckIndex; 278 WebRtcIsac_EncodeReceiveBw(&intVar, &ISACencLB_obj->bitstr_obj); 279 } 280 281 /* Split signal in two bands. */ 282 WebRtcIsac_SplitAndFilterFloat(ISACencLB_obj->data_buffer_float, LP, HP, 283 LP_lookahead, HP_lookahead, 284 &ISACencLB_obj->prefiltbankstr_obj); 285 286 /* estimate pitch parameters and pitch-filter lookahead signal */ 287 WebRtcIsac_PitchAnalysis(LP_lookahead, LP_lookahead_pf, 288 &ISACencLB_obj->pitchanalysisstr_obj, PitchLags, 289 PitchGains); 290 291 /* Encode in FIX Q12. */ 292 293 /* Convert PitchGain to Fixed point. */ 294 for (k = 0; k < PITCH_SUBFRAMES; k++) { 295 PitchGains_Q12[k] = (WebRtc_Word16)(PitchGains[k] * 4096.0); 296 } 297 298 /* Set where to store data in multiple packets memory. */ 299 if (frame_mode == 0 || ISACencLB_obj->frame_nb == 0) { 300 ISACencLB_obj->SaveEnc_obj.startIdx = 0; 301 } else { 302 ISACencLB_obj->SaveEnc_obj.startIdx = 1; 303 } 304 305 /* Quantize & encode pitch parameters. */ 306 WebRtcIsac_EncodePitchGain(PitchGains_Q12, &ISACencLB_obj->bitstr_obj, 307 &ISACencLB_obj->SaveEnc_obj); 308 WebRtcIsac_EncodePitchLag(PitchLags, PitchGains_Q12, 309 &ISACencLB_obj->bitstr_obj, 310 &ISACencLB_obj->SaveEnc_obj); 311 312 AvgPitchGain_Q12 = (PitchGains_Q12[0] + PitchGains_Q12[1] + 313 PitchGains_Q12[2] + PitchGains_Q12[3]) >> 2; 314 315 /* Find coefficients for perceptual pre-filters. */ 316 WebRtcIsac_GetLpcCoefLb(LP_lookahead_pf, HP_lookahead, 317 &ISACencLB_obj->maskfiltstr_obj, ISACencLB_obj->s2nr, 318 PitchGains_Q12, lofilt_coef, hifilt_coef); 319 320 /* Code LPC model and shape - gains not quantized yet. */ 321 WebRtcIsac_EncodeLpcLb(lofilt_coef, hifilt_coef, &ISACencLB_obj->bitstr_obj, 322 &ISACencLB_obj->SaveEnc_obj); 323 324 /* Convert PitchGains back to FLOAT for pitchfilter_pre. */ 325 for (k = 0; k < 4; k++) { 326 PitchGains[k] = ((float)PitchGains_Q12[k]) / 4096; 327 } 328 329 /* Store the state of arithmetic coder before coding LPC gains. */ 330 transcodingParam.W_upper = ISACencLB_obj->bitstr_obj.W_upper; 331 transcodingParam.stream_index = ISACencLB_obj->bitstr_obj.stream_index; 332 transcodingParam.streamval = ISACencLB_obj->bitstr_obj.streamval; 333 transcodingParam.stream[0] = 334 ISACencLB_obj->bitstr_obj.stream[ISACencLB_obj->bitstr_obj.stream_index - 335 2]; 336 transcodingParam.stream[1] = 337 ISACencLB_obj->bitstr_obj.stream[ISACencLB_obj->bitstr_obj.stream_index - 338 1]; 339 transcodingParam.stream[2] = 340 ISACencLB_obj->bitstr_obj.stream[ISACencLB_obj->bitstr_obj.stream_index]; 341 342 /* Store LPC Gains before encoding them. */ 343 for (k = 0; k < SUBFRAMES; k++) { 344 transcodingParam.loFiltGain[k] = lofilt_coef[(LPC_LOBAND_ORDER + 1) * k]; 345 transcodingParam.hiFiltGain[k] = hifilt_coef[(LPC_HIBAND_ORDER + 1) * k]; 346 } 347 348 /* Code gains */ 349 WebRtcIsac_EncodeLpcGainLb(lofilt_coef, hifilt_coef, 350 &ISACencLB_obj->bitstr_obj, 351 &ISACencLB_obj->SaveEnc_obj); 352 353 /* Get the correct value for the payload limit and calculate the 354 * number of bytes left for coding the spectrum. */ 355 if ((frame_mode == 1) && (ISACencLB_obj->frame_nb == 0)) { 356 /* It is a 60ms and we are in the first 30ms then the limit at 357 * this point should be half of the assigned value. */ 358 payloadLimitBytes = ISACencLB_obj->payloadLimitBytes60 >> 1; 359 } else if (frame_mode == 0) { 360 /* It is a 30ms frame */ 361 /* Subract 3 because termination process may add 3 bytes. */ 362 payloadLimitBytes = ISACencLB_obj->payloadLimitBytes30 - 3; 363 } else { 364 /* This is the second half of a 60ms frame. */ 365 /* Subract 3 because termination process may add 3 bytes. */ 366 payloadLimitBytes = ISACencLB_obj->payloadLimitBytes60 - 3; 367 } 368 bytesLeftSpecCoding = payloadLimitBytes - transcodingParam.stream_index; 369 370 /* Perceptual pre-filtering (using normalized lattice filter). */ 371 /* Low-band filtering. */ 372 WebRtcIsac_NormLatticeFilterMa(ORDERLO, 373 ISACencLB_obj->maskfiltstr_obj.PreStateLoF, 374 ISACencLB_obj->maskfiltstr_obj.PreStateLoG, 375 LP, lofilt_coef, LPw); 376 /* High-band filtering. */ 377 WebRtcIsac_NormLatticeFilterMa(ORDERHI, 378 ISACencLB_obj->maskfiltstr_obj.PreStateHiF, 379 ISACencLB_obj->maskfiltstr_obj.PreStateHiG, 380 HP, hifilt_coef, HPw); 381 /* Pitch filter. */ 382 WebRtcIsac_PitchfilterPre(LPw, LPw_pf, &ISACencLB_obj->pitchfiltstr_obj, 383 PitchLags, PitchGains); 384 /* Transform */ 385 WebRtcIsac_Time2Spec(LPw_pf, HPw, fre, fim, &ISACencLB_obj->fftstr_obj); 386 387 /* Save data for multiple packets memory. */ 388 my_index = ISACencLB_obj->SaveEnc_obj.startIdx * FRAMESAMPLES_HALF; 389 memcpy(&ISACencLB_obj->SaveEnc_obj.fre[my_index], fre, sizeof(fre)); 390 memcpy(&ISACencLB_obj->SaveEnc_obj.fim[my_index], fim, sizeof(fim)); 391 392 ISACencLB_obj->SaveEnc_obj.AvgPitchGain[ISACencLB_obj->SaveEnc_obj.startIdx] = 393 AvgPitchGain_Q12; 394 395 /* Quantization and loss-less coding. */ 396 err = WebRtcIsac_EncodeSpec(fre, fim, AvgPitchGain_Q12, kIsacLowerBand, 397 &ISACencLB_obj->bitstr_obj); 398 if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 399 /* There has been an error but it was not too large payload 400 (we can cure too large payload). */ 401 if (frame_mode == 1 && ISACencLB_obj->frame_nb == 1) { 402 /* If this is the second 30ms of a 60ms frame reset 403 this such that in the next call encoder starts fresh. */ 404 ISACencLB_obj->frame_nb = 0; 405 } 406 return err; 407 } 408 iterCntr = 0; 409 while ((ISACencLB_obj->bitstr_obj.stream_index > payloadLimitBytes) || 410 (err == -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 411 double bytesSpecCoderUsed; 412 double transcodeScale; 413 414 if (iterCntr >= MAX_PAYLOAD_LIMIT_ITERATION) { 415 /* We were not able to limit the payload size */ 416 if ((frame_mode == 1) && (ISACencLB_obj->frame_nb == 0)) { 417 /* This was the first 30ms of a 60ms frame. Although 418 the payload is larger than it should be but we let 419 the second 30ms be encoded. Maybe together we 420 won't exceed the limit. */ 421 ISACencLB_obj->frame_nb = 1; 422 return 0; 423 } else if ((frame_mode == 1) && (ISACencLB_obj->frame_nb == 1)) { 424 ISACencLB_obj->frame_nb = 0; 425 } 426 427 if (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH) { 428 return -ISAC_PAYLOAD_LARGER_THAN_LIMIT; 429 } else { 430 return status; 431 } 432 } 433 434 if (err == -ISAC_DISALLOWED_BITSTREAM_LENGTH) { 435 bytesSpecCoderUsed = STREAM_SIZE_MAX; 436 /* Being conservative */ 437 transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed * 0.5; 438 } else { 439 bytesSpecCoderUsed = ISACencLB_obj->bitstr_obj.stream_index - 440 transcodingParam.stream_index; 441 transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed; 442 } 443 444 /* To be safe, we reduce the scale depending on 445 the number of iterations. */ 446 transcodeScale *= (1.0 - (0.9 * (double)iterCntr / 447 (double)MAX_PAYLOAD_LIMIT_ITERATION)); 448 449 /* Scale the LPC Gains. */ 450 for (k = 0; k < SUBFRAMES; k++) { 451 lofilt_coef[(LPC_LOBAND_ORDER + 1) * k] = 452 transcodingParam.loFiltGain[k] * transcodeScale; 453 hifilt_coef[(LPC_HIBAND_ORDER + 1) * k] = 454 transcodingParam.hiFiltGain[k] * transcodeScale; 455 transcodingParam.loFiltGain[k] = lofilt_coef[(LPC_LOBAND_ORDER + 1) * k]; 456 transcodingParam.hiFiltGain[k] = hifilt_coef[(LPC_HIBAND_ORDER + 1) * k]; 457 } 458 459 /* Scale DFT coefficients. */ 460 for (k = 0; k < FRAMESAMPLES_HALF; k++) { 461 fre[k] = (WebRtc_Word16)(fre[k] * transcodeScale); 462 fim[k] = (WebRtc_Word16)(fim[k] * transcodeScale); 463 } 464 465 /* Save data for multiple packets memory. */ 466 my_index = ISACencLB_obj->SaveEnc_obj.startIdx * FRAMESAMPLES_HALF; 467 memcpy(&ISACencLB_obj->SaveEnc_obj.fre[my_index], fre, sizeof(fre)); 468 memcpy(&ISACencLB_obj->SaveEnc_obj.fim[my_index], fim, sizeof(fim)); 469 470 /* Re-store the state of arithmetic coder before coding LPC gains. */ 471 ISACencLB_obj->bitstr_obj.W_upper = transcodingParam.W_upper; 472 ISACencLB_obj->bitstr_obj.stream_index = transcodingParam.stream_index; 473 ISACencLB_obj->bitstr_obj.streamval = transcodingParam.streamval; 474 ISACencLB_obj->bitstr_obj.stream[transcodingParam.stream_index - 2] = 475 transcodingParam.stream[0]; 476 ISACencLB_obj->bitstr_obj.stream[transcodingParam.stream_index - 1] = 477 transcodingParam.stream[1]; 478 ISACencLB_obj->bitstr_obj.stream[transcodingParam.stream_index] = 479 transcodingParam.stream[2]; 480 481 /* Code gains. */ 482 WebRtcIsac_EncodeLpcGainLb(lofilt_coef, hifilt_coef, 483 &ISACencLB_obj->bitstr_obj, 484 &ISACencLB_obj->SaveEnc_obj); 485 486 /* Update the number of bytes left for encoding the spectrum. */ 487 bytesLeftSpecCoding = payloadLimitBytes - transcodingParam.stream_index; 488 489 /* Encode the spectrum. */ 490 err = WebRtcIsac_EncodeSpec(fre, fim, AvgPitchGain_Q12, kIsacLowerBand, 491 &ISACencLB_obj->bitstr_obj); 492 493 if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 494 /* There has been an error but it was not too large 495 payload (we can cure too large payload). */ 496 if (frame_mode == 1 && ISACencLB_obj->frame_nb == 1) { 497 /* If this is the second 30 ms of a 60 ms frame reset 498 this such that in the next call encoder starts fresh. */ 499 ISACencLB_obj->frame_nb = 0; 500 } 501 return err; 502 } 503 iterCntr++; 504 } 505 506 /* If 60 ms frame-size and just processed the first 30 ms, */ 507 /* go back to main function to buffer the other 30 ms speech frame. */ 508 if (frame_mode == 1) { 509 if (ISACencLB_obj->frame_nb == 0) { 510 ISACencLB_obj->frame_nb = 1; 511 return 0; 512 } else if (ISACencLB_obj->frame_nb == 1) { 513 ISACencLB_obj->frame_nb = 0; 514 /* Also update the frame-length for next packet, 515 in Adaptive mode only. */ 516 if (codingMode == 0 && (ISACencLB_obj->enforceFrameSize == 0)) { 517 ISACencLB_obj->new_framelength = 518 WebRtcIsac_GetNewFrameLength(ISACencLB_obj->bottleneck, 519 ISACencLB_obj->current_framesamples); 520 } 521 } 522 } else { 523 ISACencLB_obj->frame_nb = 0; 524 } 525 526 /* Complete arithmetic coding. */ 527 stream_length = WebRtcIsac_EncTerminate(&ISACencLB_obj->bitstr_obj); 528 return stream_length; 529 } 530 531 532 533 static int LimitPayloadUb(ISACUBEncStruct* ISACencUB_obj, 534 WebRtc_UWord16 payloadLimitBytes, 535 double bytesLeftSpecCoding, 536 transcode_obj* transcodingParam, 537 WebRtc_Word16* fre, WebRtc_Word16* fim, 538 double* lpcGains, enum ISACBand band, int status) { 539 540 int iterCntr = 0; 541 int k; 542 double bytesSpecCoderUsed; 543 double transcodeScale; 544 const WebRtc_Word16 kAveragePitchGain = 0.0; 545 546 do { 547 if (iterCntr >= MAX_PAYLOAD_LIMIT_ITERATION) { 548 /* We were not able to limit the payload size. */ 549 return -ISAC_PAYLOAD_LARGER_THAN_LIMIT; 550 } 551 552 if (status == -ISAC_DISALLOWED_BITSTREAM_LENGTH) { 553 bytesSpecCoderUsed = STREAM_SIZE_MAX; 554 /* Being conservative. */ 555 transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed * 0.5; 556 } else { 557 bytesSpecCoderUsed = ISACencUB_obj->bitstr_obj.stream_index - 558 transcodingParam->stream_index; 559 transcodeScale = bytesLeftSpecCoding / bytesSpecCoderUsed; 560 } 561 562 /* To be safe, we reduce the scale depending on the 563 number of iterations. */ 564 transcodeScale *= (1.0 - (0.9 * (double)iterCntr / 565 (double)MAX_PAYLOAD_LIMIT_ITERATION)); 566 567 /* Scale the LPC Gains. */ 568 if (band == kIsacUpperBand16) { 569 /* Two sets of coefficients if 16 kHz. */ 570 for (k = 0; k < SUBFRAMES; k++) { 571 transcodingParam->loFiltGain[k] *= transcodeScale; 572 transcodingParam->hiFiltGain[k] *= transcodeScale; 573 } 574 } else { 575 /* One sets of coefficients if 12 kHz. */ 576 for (k = 0; k < SUBFRAMES; k++) { 577 transcodingParam->loFiltGain[k] *= transcodeScale; 578 } 579 } 580 581 /* Scale DFT coefficients. */ 582 for (k = 0; k < FRAMESAMPLES_HALF; k++) { 583 fre[k] = (WebRtc_Word16)(fre[k] * transcodeScale + 0.5); 584 fim[k] = (WebRtc_Word16)(fim[k] * transcodeScale + 0.5); 585 } 586 /* Store FFT coefficients for multiple encoding. */ 587 memcpy(ISACencUB_obj->SaveEnc_obj.realFFT, fre, 588 sizeof(ISACencUB_obj->SaveEnc_obj.realFFT)); 589 memcpy(ISACencUB_obj->SaveEnc_obj.imagFFT, fim, 590 sizeof(ISACencUB_obj->SaveEnc_obj.imagFFT)); 591 592 /* Store the state of arithmetic coder before coding LPC gains */ 593 ISACencUB_obj->bitstr_obj.W_upper = transcodingParam->W_upper; 594 ISACencUB_obj->bitstr_obj.stream_index = transcodingParam->stream_index; 595 ISACencUB_obj->bitstr_obj.streamval = transcodingParam->streamval; 596 ISACencUB_obj->bitstr_obj.stream[transcodingParam->stream_index - 2] = 597 transcodingParam->stream[0]; 598 ISACencUB_obj->bitstr_obj.stream[transcodingParam->stream_index - 1] = 599 transcodingParam->stream[1]; 600 ISACencUB_obj->bitstr_obj.stream[transcodingParam->stream_index] = 601 transcodingParam->stream[2]; 602 603 /* Store the gains for multiple encoding. */ 604 memcpy(ISACencUB_obj->SaveEnc_obj.lpcGain, lpcGains, 605 SUBFRAMES * sizeof(double)); 606 /* Entropy Code lpc-gains, indices are stored for a later use.*/ 607 WebRtcIsac_EncodeLpcGainUb(transcodingParam->loFiltGain, 608 &ISACencUB_obj->bitstr_obj, 609 ISACencUB_obj->SaveEnc_obj.lpcGainIndex); 610 611 /* If 16kHz should do one more set. */ 612 if (band == kIsacUpperBand16) { 613 /* Store the gains for multiple encoding. */ 614 memcpy(&ISACencUB_obj->SaveEnc_obj.lpcGain[SUBFRAMES], 615 &lpcGains[SUBFRAMES], SUBFRAMES * sizeof(double)); 616 /* Entropy Code lpc-gains, indices are stored for a later use.*/ 617 WebRtcIsac_EncodeLpcGainUb( 618 transcodingParam->hiFiltGain, &ISACencUB_obj->bitstr_obj, 619 &ISACencUB_obj->SaveEnc_obj.lpcGainIndex[SUBFRAMES]); 620 } 621 622 /* Update the number of bytes left for encoding the spectrum. */ 623 bytesLeftSpecCoding = payloadLimitBytes - 624 ISACencUB_obj->bitstr_obj.stream_index; 625 626 /* Save the bit-stream object at this point for FEC. */ 627 memcpy(&ISACencUB_obj->SaveEnc_obj.bitStreamObj, 628 &ISACencUB_obj->bitstr_obj, sizeof(Bitstr)); 629 630 /* Encode the spectrum. */ 631 status = WebRtcIsac_EncodeSpec(fre, fim, kAveragePitchGain, 632 band, &ISACencUB_obj->bitstr_obj); 633 if ((status < 0) && (status != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 634 /* There has been an error but it was not too large payload 635 (we can cure too large payload). */ 636 return status; 637 } 638 iterCntr++; 639 } while ((ISACencUB_obj->bitstr_obj.stream_index > payloadLimitBytes) || 640 (status == -ISAC_DISALLOWED_BITSTREAM_LENGTH)); 641 return 0; 642 } 643 644 int WebRtcIsac_EncodeUb16(float* in, ISACUBEncStruct* ISACencUB_obj, 645 WebRtc_Word32 jitterInfo) { 646 int err; 647 int k; 648 649 double lpcVecs[UB_LPC_ORDER * UB16_LPC_VEC_PER_FRAME]; 650 double percepFilterParams[(1 + UB_LPC_ORDER) * (SUBFRAMES << 1) + 651 (1 + UB_LPC_ORDER)]; 652 653 double LP_lookahead[FRAMESAMPLES]; 654 WebRtc_Word16 fre[FRAMESAMPLES_HALF]; /* Q7 */ 655 WebRtc_Word16 fim[FRAMESAMPLES_HALF]; /* Q7 */ 656 657 int status = 0; 658 659 double varscale[2]; 660 double corr[SUBFRAMES << 1][UB_LPC_ORDER + 1]; 661 double lpcGains[SUBFRAMES << 1]; 662 transcode_obj transcodingParam; 663 WebRtc_UWord16 payloadLimitBytes; 664 double s2nr; 665 const WebRtc_Word16 kAveragePitchGain = 0.0; 666 int bytesLeftSpecCoding; 667 668 /* Buffer speech samples (by 10ms packet) until the frame-length is */ 669 /* reached (30 ms). */ 670 /*********************************************************************/ 671 672 /* fill the buffer with 10ms input data */ 673 memcpy(&ISACencUB_obj->data_buffer_float[ISACencUB_obj->buffer_index], in, 674 FRAMESAMPLES_10ms * sizeof(float)); 675 676 /* If buffer size is not equal to current frame-size, and end of file is 677 * not reached yet, we don't do encoding unless we have the whole frame. */ 678 if (ISACencUB_obj->buffer_index + FRAMESAMPLES_10ms < FRAMESAMPLES) { 679 ISACencUB_obj->buffer_index += FRAMESAMPLES_10ms; 680 return 0; 681 } 682 683 /* End of buffer function. */ 684 /**************************/ 685 686 /* Encoding */ 687 /************/ 688 689 /* Reset bit-stream */ 690 WebRtcIsac_ResetBitstream(&(ISACencUB_obj->bitstr_obj)); 691 692 /* Encoding of bandwidth information. */ 693 WebRtcIsac_EncodeJitterInfo(jitterInfo, &ISACencUB_obj->bitstr_obj); 694 695 status = WebRtcIsac_EncodeBandwidth(isac16kHz, &ISACencUB_obj->bitstr_obj); 696 if (status < 0) { 697 return status; 698 } 699 700 s2nr = WebRtcIsac_GetSnr(ISACencUB_obj->bottleneck, FRAMESAMPLES); 701 702 memcpy(lpcVecs, ISACencUB_obj->lastLPCVec, UB_LPC_ORDER * sizeof(double)); 703 704 for (k = 0; k < FRAMESAMPLES; k++) { 705 LP_lookahead[k] = ISACencUB_obj->data_buffer_float[UB_LOOKAHEAD + k]; 706 } 707 708 /* Find coefficients for perceptual pre-filters. */ 709 WebRtcIsac_GetLpcCoefUb(LP_lookahead, &ISACencUB_obj->maskfiltstr_obj, 710 &lpcVecs[UB_LPC_ORDER], corr, varscale, isac16kHz); 711 712 memcpy(ISACencUB_obj->lastLPCVec, 713 &lpcVecs[(UB16_LPC_VEC_PER_FRAME - 1) * (UB_LPC_ORDER)], 714 sizeof(double) * UB_LPC_ORDER); 715 716 /* Code LPC model and shape - gains not quantized yet. */ 717 WebRtcIsac_EncodeLpcUB(lpcVecs, &ISACencUB_obj->bitstr_obj, 718 percepFilterParams, isac16kHz, 719 &ISACencUB_obj->SaveEnc_obj); 720 721 /* the first set of lpc parameters are from the last sub-frame of 722 * the previous frame. so we don't care about them. */ 723 WebRtcIsac_GetLpcGain(s2nr, &percepFilterParams[UB_LPC_ORDER + 1], 724 (SUBFRAMES << 1), lpcGains, corr, varscale); 725 726 /* Store the state of arithmetic coder before coding LPC gains */ 727 transcodingParam.stream_index = ISACencUB_obj->bitstr_obj.stream_index; 728 transcodingParam.W_upper = ISACencUB_obj->bitstr_obj.W_upper; 729 transcodingParam.streamval = ISACencUB_obj->bitstr_obj.streamval; 730 transcodingParam.stream[0] = 731 ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index - 732 2]; 733 transcodingParam.stream[1] = 734 ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index - 735 1]; 736 transcodingParam.stream[2] = 737 ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index]; 738 739 /* Store LPC Gains before encoding them. */ 740 for (k = 0; k < SUBFRAMES; k++) { 741 transcodingParam.loFiltGain[k] = lpcGains[k]; 742 transcodingParam.hiFiltGain[k] = lpcGains[SUBFRAMES + k]; 743 } 744 745 /* Store the gains for multiple encoding. */ 746 memcpy(ISACencUB_obj->SaveEnc_obj.lpcGain, lpcGains, 747 (SUBFRAMES << 1) * sizeof(double)); 748 749 WebRtcIsac_EncodeLpcGainUb(lpcGains, &ISACencUB_obj->bitstr_obj, 750 ISACencUB_obj->SaveEnc_obj.lpcGainIndex); 751 WebRtcIsac_EncodeLpcGainUb( 752 &lpcGains[SUBFRAMES], &ISACencUB_obj->bitstr_obj, 753 &ISACencUB_obj->SaveEnc_obj.lpcGainIndex[SUBFRAMES]); 754 755 /* Get the correct value for the payload limit and calculate the number of 756 bytes left for coding the spectrum. It is a 30ms frame 757 Subract 3 because termination process may add 3 bytes */ 758 payloadLimitBytes = ISACencUB_obj->maxPayloadSizeBytes - 759 ISACencUB_obj->numBytesUsed - 3; 760 bytesLeftSpecCoding = payloadLimitBytes - 761 ISACencUB_obj->bitstr_obj.stream_index; 762 763 for (k = 0; k < (SUBFRAMES << 1); k++) { 764 percepFilterParams[k * (UB_LPC_ORDER + 1) + (UB_LPC_ORDER + 1)] = 765 lpcGains[k]; 766 } 767 768 /* LPC filtering (using normalized lattice filter), */ 769 /* first half-frame. */ 770 WebRtcIsac_NormLatticeFilterMa(UB_LPC_ORDER, 771 ISACencUB_obj->maskfiltstr_obj.PreStateLoF, 772 ISACencUB_obj->maskfiltstr_obj.PreStateLoG, 773 &ISACencUB_obj->data_buffer_float[0], 774 &percepFilterParams[UB_LPC_ORDER + 1], 775 &LP_lookahead[0]); 776 777 /* Second half-frame filtering. */ 778 WebRtcIsac_NormLatticeFilterMa( 779 UB_LPC_ORDER, ISACencUB_obj->maskfiltstr_obj.PreStateLoF, 780 ISACencUB_obj->maskfiltstr_obj.PreStateLoG, 781 &ISACencUB_obj->data_buffer_float[FRAMESAMPLES_HALF], 782 &percepFilterParams[(UB_LPC_ORDER + 1) + SUBFRAMES * (UB_LPC_ORDER + 1)], 783 &LP_lookahead[FRAMESAMPLES_HALF]); 784 785 WebRtcIsac_Time2Spec(&LP_lookahead[0], &LP_lookahead[FRAMESAMPLES_HALF], 786 fre, fim, &ISACencUB_obj->fftstr_obj); 787 788 /* Store FFT coefficients for multiple encoding. */ 789 memcpy(ISACencUB_obj->SaveEnc_obj.realFFT, fre, sizeof(fre)); 790 memcpy(ISACencUB_obj->SaveEnc_obj.imagFFT, fim, sizeof(fim)); 791 792 /* Prepare the audio buffer for the next packet 793 * move the last 3 ms to the beginning of the buffer. */ 794 memcpy(ISACencUB_obj->data_buffer_float, 795 &ISACencUB_obj->data_buffer_float[FRAMESAMPLES], 796 LB_TOTAL_DELAY_SAMPLES * sizeof(float)); 797 /* start writing with 3 ms delay to compensate for the delay 798 * of the lower-band. */ 799 ISACencUB_obj->buffer_index = LB_TOTAL_DELAY_SAMPLES; 800 801 /* Save the bit-stream object at this point for FEC. */ 802 memcpy(&ISACencUB_obj->SaveEnc_obj.bitStreamObj, &ISACencUB_obj->bitstr_obj, 803 sizeof(Bitstr)); 804 805 /* Qantization and lossless coding */ 806 /* Note that there is no pitch-gain for this band so kAveragePitchGain = 0 807 * is passed to the function. In fact, the function ignores the 3rd parameter 808 * for this band. */ 809 err = WebRtcIsac_EncodeSpec(fre, fim, kAveragePitchGain, kIsacUpperBand16, 810 &ISACencUB_obj->bitstr_obj); 811 if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 812 return err; 813 } 814 815 if ((ISACencUB_obj->bitstr_obj.stream_index > payloadLimitBytes) || 816 (err == -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 817 err = LimitPayloadUb(ISACencUB_obj, payloadLimitBytes, bytesLeftSpecCoding, 818 &transcodingParam, fre, fim, lpcGains, 819 kIsacUpperBand16, err); 820 } 821 if (err < 0) { 822 return err; 823 } 824 /* Complete arithmetic coding. */ 825 return WebRtcIsac_EncTerminate(&ISACencUB_obj->bitstr_obj); 826 } 827 828 829 int WebRtcIsac_EncodeUb12(float* in, ISACUBEncStruct* ISACencUB_obj, 830 WebRtc_Word32 jitterInfo) { 831 int err; 832 int k; 833 834 double lpcVecs[UB_LPC_ORDER * UB_LPC_VEC_PER_FRAME]; 835 836 double percepFilterParams[(1 + UB_LPC_ORDER) * SUBFRAMES]; 837 float LP[FRAMESAMPLES_HALF]; 838 float HP[FRAMESAMPLES_HALF]; 839 840 double LP_lookahead[FRAMESAMPLES_HALF]; 841 double HP_lookahead[FRAMESAMPLES_HALF]; 842 double LPw[FRAMESAMPLES_HALF]; 843 844 double HPw[FRAMESAMPLES_HALF]; 845 WebRtc_Word16 fre[FRAMESAMPLES_HALF]; /* Q7 */ 846 WebRtc_Word16 fim[FRAMESAMPLES_HALF]; /* Q7 */ 847 848 int status = 0; 849 850 double varscale[1]; 851 852 double corr[UB_LPC_GAIN_DIM][UB_LPC_ORDER + 1]; 853 double lpcGains[SUBFRAMES]; 854 transcode_obj transcodingParam; 855 WebRtc_UWord16 payloadLimitBytes; 856 double s2nr; 857 const WebRtc_Word16 kAveragePitchGain = 0.0; 858 double bytesLeftSpecCoding; 859 860 /* Buffer speech samples (by 10ms packet) until the framelength is */ 861 /* reached (30 ms). */ 862 /********************************************************************/ 863 864 /* Fill the buffer with 10ms input data. */ 865 memcpy(&ISACencUB_obj->data_buffer_float[ISACencUB_obj->buffer_index], in, 866 FRAMESAMPLES_10ms * sizeof(float)); 867 868 /* if buffer-size is not equal to current frame-size then increase the 869 index and return. We do the encoding when we have enough audio. */ 870 if (ISACencUB_obj->buffer_index + FRAMESAMPLES_10ms < FRAMESAMPLES) { 871 ISACencUB_obj->buffer_index += FRAMESAMPLES_10ms; 872 return 0; 873 } 874 /* If buffer reached the right size, reset index and continue 875 with encoding the frame */ 876 ISACencUB_obj->buffer_index = 0; 877 878 /* End of buffer function */ 879 /**************************/ 880 881 /* Encoding */ 882 /************/ 883 884 /* Reset bit-stream. */ 885 WebRtcIsac_ResetBitstream(&(ISACencUB_obj->bitstr_obj)); 886 887 /* Encoding bandwidth information. */ 888 WebRtcIsac_EncodeJitterInfo(jitterInfo, &ISACencUB_obj->bitstr_obj); 889 status = WebRtcIsac_EncodeBandwidth(isac12kHz, &ISACencUB_obj->bitstr_obj); 890 if (status < 0) { 891 return status; 892 } 893 894 s2nr = WebRtcIsac_GetSnr(ISACencUB_obj->bottleneck, FRAMESAMPLES); 895 896 /* Split signal in two bands. */ 897 WebRtcIsac_SplitAndFilterFloat(ISACencUB_obj->data_buffer_float, HP, LP, 898 HP_lookahead, LP_lookahead, 899 &ISACencUB_obj->prefiltbankstr_obj); 900 901 /* Find coefficients for perceptual pre-filters. */ 902 WebRtcIsac_GetLpcCoefUb(LP_lookahead, &ISACencUB_obj->maskfiltstr_obj, 903 lpcVecs, corr, varscale, isac12kHz); 904 905 /* Code LPC model and shape - gains not quantized yet. */ 906 WebRtcIsac_EncodeLpcUB(lpcVecs, &ISACencUB_obj->bitstr_obj, 907 percepFilterParams, isac12kHz, 908 &ISACencUB_obj->SaveEnc_obj); 909 910 WebRtcIsac_GetLpcGain(s2nr, percepFilterParams, SUBFRAMES, lpcGains, corr, 911 varscale); 912 913 /* Store the state of arithmetic coder before coding LPC gains. */ 914 transcodingParam.W_upper = ISACencUB_obj->bitstr_obj.W_upper; 915 transcodingParam.stream_index = ISACencUB_obj->bitstr_obj.stream_index; 916 transcodingParam.streamval = ISACencUB_obj->bitstr_obj.streamval; 917 transcodingParam.stream[0] = 918 ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index - 919 2]; 920 transcodingParam.stream[1] = 921 ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index - 922 1]; 923 transcodingParam.stream[2] = 924 ISACencUB_obj->bitstr_obj.stream[ISACencUB_obj->bitstr_obj.stream_index]; 925 926 /* Store LPC Gains before encoding them. */ 927 for (k = 0; k < SUBFRAMES; k++) { 928 transcodingParam.loFiltGain[k] = lpcGains[k]; 929 } 930 931 /* Store the gains for multiple encoding. */ 932 memcpy(ISACencUB_obj->SaveEnc_obj.lpcGain, lpcGains, SUBFRAMES * 933 sizeof(double)); 934 935 WebRtcIsac_EncodeLpcGainUb(lpcGains, &ISACencUB_obj->bitstr_obj, 936 ISACencUB_obj->SaveEnc_obj.lpcGainIndex); 937 938 for (k = 0; k < SUBFRAMES; k++) { 939 percepFilterParams[k * (UB_LPC_ORDER + 1)] = lpcGains[k]; 940 } 941 942 /* perceptual pre-filtering (using normalized lattice filter) */ 943 /* low-band filtering */ 944 WebRtcIsac_NormLatticeFilterMa(UB_LPC_ORDER, 945 ISACencUB_obj->maskfiltstr_obj.PreStateLoF, 946 ISACencUB_obj->maskfiltstr_obj.PreStateLoG, LP, 947 percepFilterParams, LPw); 948 949 /* Get the correct value for the payload limit and calculate the number 950 of bytes left for coding the spectrum. It is a 30ms frame Subract 3 951 because termination process may add 3 bytes */ 952 payloadLimitBytes = ISACencUB_obj->maxPayloadSizeBytes - 953 ISACencUB_obj->numBytesUsed - 3; 954 bytesLeftSpecCoding = payloadLimitBytes - 955 ISACencUB_obj->bitstr_obj.stream_index; 956 957 memset(HPw, 0, sizeof(HPw)); 958 959 /* Transform */ 960 WebRtcIsac_Time2Spec(LPw, HPw, fre, fim, &ISACencUB_obj->fftstr_obj); 961 962 /* Store FFT coefficients for multiple encoding. */ 963 memcpy(ISACencUB_obj->SaveEnc_obj.realFFT, fre, 964 sizeof(ISACencUB_obj->SaveEnc_obj.realFFT)); 965 memcpy(ISACencUB_obj->SaveEnc_obj.imagFFT, fim, 966 sizeof(ISACencUB_obj->SaveEnc_obj.imagFFT)); 967 968 /* Save the bit-stream object at this point for FEC. */ 969 memcpy(&ISACencUB_obj->SaveEnc_obj.bitStreamObj, 970 &ISACencUB_obj->bitstr_obj, sizeof(Bitstr)); 971 972 /* Quantization and loss-less coding */ 973 /* The 4th parameter to this function is pitch-gain, which is only used 974 * when encoding 0-8 kHz band, and irrelevant in this function, therefore, 975 * we insert zero here. */ 976 err = WebRtcIsac_EncodeSpec(fre, fim, kAveragePitchGain, kIsacUpperBand12, 977 &ISACencUB_obj->bitstr_obj); 978 if ((err < 0) && (err != -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 979 /* There has been an error but it was not too large 980 payload (we can cure too large payload) */ 981 return err; 982 } 983 984 if ((ISACencUB_obj->bitstr_obj.stream_index > payloadLimitBytes) || 985 (err == -ISAC_DISALLOWED_BITSTREAM_LENGTH)) { 986 err = LimitPayloadUb(ISACencUB_obj, payloadLimitBytes, bytesLeftSpecCoding, 987 &transcodingParam, fre, fim, lpcGains, 988 kIsacUpperBand12, err); 989 } 990 if (err < 0) { 991 return err; 992 } 993 /* Complete arithmetic coding. */ 994 return WebRtcIsac_EncTerminate(&ISACencUB_obj->bitstr_obj); 995 } 996 997 998 999 1000 1001 1002 /* This function is used to create a new bit-stream with new BWE. 1003 The same data as previously encoded with the function WebRtcIsac_Encoder(). 1004 The data needed is taken from the structure, where it was stored 1005 when calling the encoder. */ 1006 1007 int WebRtcIsac_EncodeStoredDataLb(const ISAC_SaveEncData_t* ISACSavedEnc_obj, 1008 Bitstr* ISACBitStr_obj, int BWnumber, 1009 float scale) { 1010 int ii; 1011 int status; 1012 int BWno = BWnumber; 1013 1014 const WebRtc_UWord16* WebRtcIsac_kQPitchGainCdf_ptr[1]; 1015 const WebRtc_UWord16** cdf; 1016 1017 double tmpLPCcoeffs_lo[(ORDERLO + 1)*SUBFRAMES * 2]; 1018 double tmpLPCcoeffs_hi[(ORDERHI + 1)*SUBFRAMES * 2]; 1019 int tmpLPCindex_g[12 * 2]; 1020 WebRtc_Word16 tmp_fre[FRAMESAMPLES], tmp_fim[FRAMESAMPLES]; 1021 const int kModel = 0; 1022 1023 /* Sanity Check - possible values for BWnumber is 0 - 23. */ 1024 if ((BWnumber < 0) || (BWnumber > 23)) { 1025 return -ISAC_RANGE_ERROR_BW_ESTIMATOR; 1026 } 1027 1028 /* Reset bit-stream. */ 1029 WebRtcIsac_ResetBitstream(ISACBitStr_obj); 1030 1031 /* Encode frame length */ 1032 status = WebRtcIsac_EncodeFrameLen(ISACSavedEnc_obj->framelength, 1033 ISACBitStr_obj); 1034 if (status < 0) { 1035 /* Wrong frame size. */ 1036 return status; 1037 } 1038 1039 /* Transcoding */ 1040 if ((scale > 0.0) && (scale < 1.0)) { 1041 /* Compensate LPC gain. */ 1042 for (ii = 0; 1043 ii < ((ORDERLO + 1)* SUBFRAMES * (1 + ISACSavedEnc_obj->startIdx)); 1044 ii++) { 1045 tmpLPCcoeffs_lo[ii] = scale * ISACSavedEnc_obj->LPCcoeffs_lo[ii]; 1046 } 1047 for (ii = 0; 1048 ii < ((ORDERHI + 1) * SUBFRAMES * (1 + ISACSavedEnc_obj->startIdx)); 1049 ii++) { 1050 tmpLPCcoeffs_hi[ii] = scale * ISACSavedEnc_obj->LPCcoeffs_hi[ii]; 1051 } 1052 /* Scale DFT. */ 1053 for (ii = 0; 1054 ii < (FRAMESAMPLES_HALF * (1 + ISACSavedEnc_obj->startIdx)); 1055 ii++) { 1056 tmp_fre[ii] = (WebRtc_Word16)((scale) * (float)ISACSavedEnc_obj->fre[ii]); 1057 tmp_fim[ii] = (WebRtc_Word16)((scale) * (float)ISACSavedEnc_obj->fim[ii]); 1058 } 1059 } else { 1060 for (ii = 0; 1061 ii < (KLT_ORDER_GAIN * (1 + ISACSavedEnc_obj->startIdx)); 1062 ii++) { 1063 tmpLPCindex_g[ii] = ISACSavedEnc_obj->LPCindex_g[ii]; 1064 } 1065 for (ii = 0; 1066 ii < (FRAMESAMPLES_HALF * (1 + ISACSavedEnc_obj->startIdx)); 1067 ii++) { 1068 tmp_fre[ii] = ISACSavedEnc_obj->fre[ii]; 1069 tmp_fim[ii] = ISACSavedEnc_obj->fim[ii]; 1070 } 1071 } 1072 1073 /* Encode bandwidth estimate. */ 1074 WebRtcIsac_EncodeReceiveBw(&BWno, ISACBitStr_obj); 1075 1076 /* Loop over number of 30 msec */ 1077 for (ii = 0; ii <= ISACSavedEnc_obj->startIdx; ii++) { 1078 /* Encode pitch gains. */ 1079 *WebRtcIsac_kQPitchGainCdf_ptr = WebRtcIsac_kQPitchGainCdf; 1080 WebRtcIsac_EncHistMulti(ISACBitStr_obj, 1081 &ISACSavedEnc_obj->pitchGain_index[ii], 1082 WebRtcIsac_kQPitchGainCdf_ptr, 1); 1083 1084 /* Entropy coding of quantization pitch lags */ 1085 /* Voicing classification. */ 1086 if (ISACSavedEnc_obj->meanGain[ii] < 0.2) { 1087 cdf = WebRtcIsac_kQPitchLagCdfPtrLo; 1088 } else if (ISACSavedEnc_obj->meanGain[ii] < 0.4) { 1089 cdf = WebRtcIsac_kQPitchLagCdfPtrMid; 1090 } else { 1091 cdf = WebRtcIsac_kQPitchLagCdfPtrHi; 1092 } 1093 WebRtcIsac_EncHistMulti(ISACBitStr_obj, 1094 &ISACSavedEnc_obj->pitchIndex[PITCH_SUBFRAMES * ii], 1095 cdf, PITCH_SUBFRAMES); 1096 1097 /* LPC */ 1098 /* Only one model exists. The entropy coding is done only for backward 1099 * compatibility. */ 1100 WebRtcIsac_EncHistMulti(ISACBitStr_obj, &kModel, 1101 WebRtcIsac_kQKltModelCdfPtr, 1); 1102 /* Entropy coding of quantization indices - LPC shape only. */ 1103 WebRtcIsac_EncHistMulti(ISACBitStr_obj, 1104 &ISACSavedEnc_obj->LPCindex_s[KLT_ORDER_SHAPE * ii], 1105 WebRtcIsac_kQKltCdfPtrShape, 1106 KLT_ORDER_SHAPE); 1107 1108 /* If transcoding, get new LPC gain indices */ 1109 if (scale < 1.0) { 1110 WebRtcIsac_TranscodeLPCCoef( 1111 &tmpLPCcoeffs_lo[(ORDERLO + 1) * SUBFRAMES * ii], 1112 &tmpLPCcoeffs_hi[(ORDERHI + 1)*SUBFRAMES * ii], 1113 &tmpLPCindex_g[KLT_ORDER_GAIN * ii]); 1114 } 1115 1116 /* Entropy coding of quantization indices - LPC gain. */ 1117 WebRtcIsac_EncHistMulti(ISACBitStr_obj, &tmpLPCindex_g[KLT_ORDER_GAIN * ii], 1118 WebRtcIsac_kQKltCdfPtrGain, KLT_ORDER_GAIN); 1119 1120 /* Quantization and loss-less coding. */ 1121 status = WebRtcIsac_EncodeSpec(&tmp_fre[ii * FRAMESAMPLES_HALF], 1122 &tmp_fim[ii * FRAMESAMPLES_HALF], 1123 ISACSavedEnc_obj->AvgPitchGain[ii], 1124 kIsacLowerBand, ISACBitStr_obj); 1125 if (status < 0) { 1126 return status; 1127 } 1128 } 1129 /* Complete arithmetic coding. */ 1130 return WebRtcIsac_EncTerminate(ISACBitStr_obj); 1131 } 1132 1133 1134 int WebRtcIsac_EncodeStoredDataUb( 1135 const ISACUBSaveEncDataStruct* ISACSavedEnc_obj, 1136 Bitstr* bitStream, 1137 WebRtc_Word32 jitterInfo, 1138 float scale, 1139 enum ISACBandwidth bandwidth) { 1140 int n; 1141 int err; 1142 double lpcGain[SUBFRAMES]; 1143 WebRtc_Word16 realFFT[FRAMESAMPLES_HALF]; 1144 WebRtc_Word16 imagFFT[FRAMESAMPLES_HALF]; 1145 const WebRtc_UWord16** shape_cdf; 1146 int shape_len; 1147 const WebRtc_Word16 kAveragePitchGain = 0.0; 1148 enum ISACBand band; 1149 /* Reset bitstream. */ 1150 WebRtcIsac_ResetBitstream(bitStream); 1151 1152 /* Encode jitter index. */ 1153 WebRtcIsac_EncodeJitterInfo(jitterInfo, bitStream); 1154 1155 err = WebRtcIsac_EncodeBandwidth(bandwidth, bitStream); 1156 if (err < 0) { 1157 return err; 1158 } 1159 1160 /* Encode LPC-shape. */ 1161 if (bandwidth == isac12kHz) { 1162 shape_cdf = WebRtcIsac_kLpcShapeCdfMatUb12; 1163 shape_len = UB_LPC_ORDER * UB_LPC_VEC_PER_FRAME; 1164 band = kIsacUpperBand12; 1165 } else { 1166 shape_cdf = WebRtcIsac_kLpcShapeCdfMatUb16; 1167 shape_len = UB_LPC_ORDER * UB16_LPC_VEC_PER_FRAME; 1168 band = kIsacUpperBand16; 1169 } 1170 WebRtcIsac_EncHistMulti(bitStream, ISACSavedEnc_obj->indexLPCShape, 1171 shape_cdf, shape_len); 1172 1173 if ((scale <= 0.0) || (scale >= 1.0)) { 1174 /* We only consider scales between zero and one. */ 1175 WebRtcIsac_EncHistMulti(bitStream, ISACSavedEnc_obj->lpcGainIndex, 1176 WebRtcIsac_kLpcGainCdfMat, UB_LPC_GAIN_DIM); 1177 if (bandwidth == isac16kHz) { 1178 /* Store gain indices of the second half. */ 1179 WebRtcIsac_EncHistMulti(bitStream, 1180 &ISACSavedEnc_obj->lpcGainIndex[SUBFRAMES], 1181 WebRtcIsac_kLpcGainCdfMat, UB_LPC_GAIN_DIM); 1182 } 1183 /* Store FFT coefficients. */ 1184 err = WebRtcIsac_EncodeSpec(ISACSavedEnc_obj->realFFT, 1185 ISACSavedEnc_obj->imagFFT, kAveragePitchGain, 1186 band, bitStream); 1187 } else { 1188 /* Scale LPC gain and FFT coefficients. */ 1189 for (n = 0; n < SUBFRAMES; n++) { 1190 lpcGain[n] = scale * ISACSavedEnc_obj->lpcGain[n]; 1191 } 1192 /* Store LPC gains. */ 1193 WebRtcIsac_StoreLpcGainUb(lpcGain, bitStream); 1194 1195 if (bandwidth == isac16kHz) { 1196 /* Scale and code the gains of the second half of the frame, if 16kHz. */ 1197 for (n = 0; n < SUBFRAMES; n++) { 1198 lpcGain[n] = scale * ISACSavedEnc_obj->lpcGain[n + SUBFRAMES]; 1199 } 1200 WebRtcIsac_StoreLpcGainUb(lpcGain, bitStream); 1201 } 1202 1203 for (n = 0; n < FRAMESAMPLES_HALF; n++) { 1204 realFFT[n] = (WebRtc_Word16)(scale * (float)ISACSavedEnc_obj->realFFT[n] + 1205 0.5f); 1206 imagFFT[n] = (WebRtc_Word16)(scale * (float)ISACSavedEnc_obj->imagFFT[n] + 1207 0.5f); 1208 } 1209 /* Store FFT coefficients. */ 1210 err = WebRtcIsac_EncodeSpec(realFFT, imagFFT, kAveragePitchGain, 1211 band, bitStream); 1212 } 1213 if (err < 0) { 1214 /* Error happened while encoding FFT coefficients. */ 1215 return err; 1216 } 1217 1218 /* Complete arithmetic coding. */ 1219 return WebRtcIsac_EncTerminate(bitStream); 1220 } 1221 1222 WebRtc_Word16 WebRtcIsac_GetRedPayloadUb( 1223 const ISACUBSaveEncDataStruct* ISACSavedEncObj, 1224 Bitstr* bitStreamObj, 1225 enum ISACBandwidth bandwidth) { 1226 int n; 1227 WebRtc_Word16 status; 1228 WebRtc_Word16 realFFT[FRAMESAMPLES_HALF]; 1229 WebRtc_Word16 imagFFT[FRAMESAMPLES_HALF]; 1230 enum ISACBand band; 1231 const WebRtc_Word16 kAveragePitchGain = 0.0; 1232 /* Store bit-stream object. */ 1233 memcpy(bitStreamObj, &ISACSavedEncObj->bitStreamObj, sizeof(Bitstr)); 1234 1235 /* Scale FFT coefficients. */ 1236 for (n = 0; n < FRAMESAMPLES_HALF; n++) { 1237 realFFT[n] = (WebRtc_Word16)((float)ISACSavedEncObj->realFFT[n] * 1238 RCU_TRANSCODING_SCALE_UB + 0.5); 1239 imagFFT[n] = (WebRtc_Word16)((float)ISACSavedEncObj->imagFFT[n] * 1240 RCU_TRANSCODING_SCALE_UB + 0.5); 1241 } 1242 1243 band = (bandwidth == isac12kHz) ? kIsacUpperBand12 : kIsacUpperBand16; 1244 status = WebRtcIsac_EncodeSpec(realFFT, imagFFT, kAveragePitchGain, band, 1245 bitStreamObj); 1246 if (status < 0) { 1247 return status; 1248 } else { 1249 /* Terminate entropy coding */ 1250 return WebRtcIsac_EncTerminate(bitStreamObj); 1251 } 1252 } 1253
