OLD | NEW |
(Empty) | |
| 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 #include "webrtc/modules/audio_coding/codecs/cng/webrtc_cng.h" |
| 12 |
| 13 #include <algorithm> |
| 14 |
| 15 #include "webrtc/common_audio/signal_processing/include/signal_processing_librar
y.h" |
| 16 |
| 17 namespace webrtc { |
| 18 |
| 19 namespace { |
| 20 |
| 21 const size_t kCngMaxOutsizeOrder = 640; |
| 22 |
| 23 // TODO(ossu): Rename the left-over WebRtcCng according to style guide. |
| 24 void WebRtcCng_K2a16(int16_t* k, int useOrder, int16_t* a); |
| 25 |
| 26 const int32_t WebRtcCng_kDbov[94] = { |
| 27 1081109975, 858756178, 682134279, 541838517, 430397633, 341876992, |
| 28 271562548, 215709799, 171344384, 136103682, 108110997, 85875618, |
| 29 68213428, 54183852, 43039763, 34187699, 27156255, 21570980, |
| 30 17134438, 13610368, 10811100, 8587562, 6821343, 5418385, |
| 31 4303976, 3418770, 2715625, 2157098, 1713444, 1361037, |
| 32 1081110, 858756, 682134, 541839, 430398, 341877, |
| 33 271563, 215710, 171344, 136104, 108111, 85876, |
| 34 68213, 54184, 43040, 34188, 27156, 21571, |
| 35 17134, 13610, 10811, 8588, 6821, 5418, |
| 36 4304, 3419, 2716, 2157, 1713, 1361, |
| 37 1081, 859, 682, 542, 430, 342, |
| 38 272, 216, 171, 136, 108, 86, |
| 39 68, 54, 43, 34, 27, 22, |
| 40 17, 14, 11, 9, 7, 5, |
| 41 4, 3, 3, 2, 2, 1, |
| 42 1, 1, 1, 1 |
| 43 }; |
| 44 |
| 45 const int16_t WebRtcCng_kCorrWindow[WEBRTC_CNG_MAX_LPC_ORDER] = { |
| 46 32702, 32636, 32570, 32505, 32439, 32374, |
| 47 32309, 32244, 32179, 32114, 32049, 31985 |
| 48 }; |
| 49 |
| 50 } // namespace |
| 51 |
| 52 ComfortNoiseDecoder::ComfortNoiseDecoder() { |
| 53 /* Needed to get the right function pointers in SPLIB. */ |
| 54 WebRtcSpl_Init(); |
| 55 Reset(); |
| 56 } |
| 57 |
| 58 void ComfortNoiseDecoder::Reset() { |
| 59 dec_seed_ = 7777; /* For debugging only. */ |
| 60 dec_target_energy_ = 0; |
| 61 dec_used_energy_ = 0; |
| 62 for (auto& c : dec_target_reflCoefs_) |
| 63 c = 0; |
| 64 for (auto& c : dec_used_reflCoefs_) |
| 65 c = 0; |
| 66 for (auto& c : dec_filtstate_) |
| 67 c = 0; |
| 68 for (auto& c : dec_filtstateLow_) |
| 69 c = 0; |
| 70 dec_order_ = 5; |
| 71 dec_target_scale_factor_ = 0; |
| 72 dec_used_scale_factor_ = 0; |
| 73 } |
| 74 |
| 75 void ComfortNoiseDecoder::UpdateSid(rtc::ArrayView<const uint8_t> sid) { |
| 76 int16_t refCs[WEBRTC_CNG_MAX_LPC_ORDER]; |
| 77 int32_t targetEnergy; |
| 78 size_t length = sid.size(); |
| 79 /* Throw away reflection coefficients of higher order than we can handle. */ |
| 80 if (length > (WEBRTC_CNG_MAX_LPC_ORDER + 1)) |
| 81 length = WEBRTC_CNG_MAX_LPC_ORDER + 1; |
| 82 |
| 83 dec_order_ = static_cast<uint16_t>(length - 1); |
| 84 |
| 85 uint8_t sid0 = std::min<uint8_t>(sid[0], 93); |
| 86 targetEnergy = WebRtcCng_kDbov[sid0]; |
| 87 /* Take down target energy to 75%. */ |
| 88 targetEnergy = targetEnergy >> 1; |
| 89 targetEnergy += targetEnergy >> 2; |
| 90 |
| 91 dec_target_energy_ = targetEnergy; |
| 92 |
| 93 /* Reconstruct coeffs with tweak for WebRtc implementation of RFC3389. */ |
| 94 if (dec_order_ == WEBRTC_CNG_MAX_LPC_ORDER) { |
| 95 for (size_t i = 0; i < (dec_order_); i++) { |
| 96 refCs[i] = sid[i + 1] << 8; /* Q7 to Q15*/ |
| 97 dec_target_reflCoefs_[i] = refCs[i]; |
| 98 } |
| 99 } else { |
| 100 for (size_t i = 0; i < (dec_order_); i++) { |
| 101 refCs[i] = (sid[i + 1] - 127) << 8; /* Q7 to Q15. */ |
| 102 dec_target_reflCoefs_[i] = refCs[i]; |
| 103 } |
| 104 } |
| 105 |
| 106 for (size_t i = (dec_order_); i < WEBRTC_CNG_MAX_LPC_ORDER; i++) { |
| 107 refCs[i] = 0; |
| 108 dec_target_reflCoefs_[i] = refCs[i]; |
| 109 } |
| 110 } |
| 111 |
| 112 bool ComfortNoiseDecoder::Generate(rtc::ArrayView<int16_t> out_data, |
| 113 bool new_period) { |
| 114 int16_t excitation[kCngMaxOutsizeOrder]; |
| 115 int16_t low[kCngMaxOutsizeOrder]; |
| 116 int16_t lpPoly[WEBRTC_CNG_MAX_LPC_ORDER + 1]; |
| 117 int16_t ReflBetaStd = 26214; /* 0.8 in q15. */ |
| 118 int16_t ReflBetaCompStd = 6553; /* 0.2 in q15. */ |
| 119 int16_t ReflBetaNewP = 19661; /* 0.6 in q15. */ |
| 120 int16_t ReflBetaCompNewP = 13107; /* 0.4 in q15. */ |
| 121 int16_t Beta, BetaC, tmp1, tmp2, tmp3; |
| 122 int32_t targetEnergy; |
| 123 int16_t En; |
| 124 int16_t temp16; |
| 125 const size_t num_samples = out_data.size(); |
| 126 |
| 127 if (num_samples > kCngMaxOutsizeOrder) { |
| 128 return false; |
| 129 } |
| 130 |
| 131 if (new_period) { |
| 132 dec_used_scale_factor_ = dec_target_scale_factor_; |
| 133 Beta = ReflBetaNewP; |
| 134 BetaC = ReflBetaCompNewP; |
| 135 } else { |
| 136 Beta = ReflBetaStd; |
| 137 BetaC = ReflBetaCompStd; |
| 138 } |
| 139 |
| 140 /* Here we use a 0.5 weighting, should possibly be modified to 0.6. */ |
| 141 tmp1 = dec_used_scale_factor_ << 2; /* Q13->Q15 */ |
| 142 tmp2 = dec_target_scale_factor_ << 2; /* Q13->Q15 */ |
| 143 tmp3 = (int16_t) WEBRTC_SPL_MUL_16_16_RSFT(tmp1, Beta, 15); |
| 144 tmp3 += (int16_t) WEBRTC_SPL_MUL_16_16_RSFT(tmp2, BetaC, 15); |
| 145 dec_used_scale_factor_ = tmp3 >> 2; /* Q15->Q13 */ |
| 146 |
| 147 dec_used_energy_ = dec_used_energy_ >> 1; |
| 148 dec_used_energy_ += dec_target_energy_ >> 1; |
| 149 |
| 150 /* Do the same for the reflection coeffs. */ |
| 151 for (size_t i = 0; i < WEBRTC_CNG_MAX_LPC_ORDER; i++) { |
| 152 dec_used_reflCoefs_[i] = (int16_t) WEBRTC_SPL_MUL_16_16_RSFT( |
| 153 dec_used_reflCoefs_[i], Beta, 15); |
| 154 dec_used_reflCoefs_[i] += (int16_t) WEBRTC_SPL_MUL_16_16_RSFT( |
| 155 dec_target_reflCoefs_[i], BetaC, 15); |
| 156 } |
| 157 |
| 158 /* Compute the polynomial coefficients. */ |
| 159 WebRtcCng_K2a16(dec_used_reflCoefs_, WEBRTC_CNG_MAX_LPC_ORDER, lpPoly); |
| 160 |
| 161 |
| 162 targetEnergy = dec_used_energy_; |
| 163 |
| 164 /* Calculate scaling factor based on filter energy. */ |
| 165 En = 8192; /* 1.0 in Q13. */ |
| 166 for (size_t i = 0; i < (WEBRTC_CNG_MAX_LPC_ORDER); i++) { |
| 167 /* Floating point value for reference. |
| 168 E *= 1.0 - (dec_used_reflCoefs_[i] / 32768.0) * |
| 169 (dec_used_reflCoefs_[i] / 32768.0); |
| 170 */ |
| 171 |
| 172 /* Same in fixed point. */ |
| 173 /* K(i).^2 in Q15. */ |
| 174 temp16 = (int16_t) WEBRTC_SPL_MUL_16_16_RSFT( |
| 175 dec_used_reflCoefs_[i], dec_used_reflCoefs_[i], 15); |
| 176 /* 1 - K(i).^2 in Q15. */ |
| 177 temp16 = 0x7fff - temp16; |
| 178 En = (int16_t) WEBRTC_SPL_MUL_16_16_RSFT(En, temp16, 15); |
| 179 } |
| 180 |
| 181 /* float scaling= sqrt(E * dec_target_energy_ / (1 << 24)); */ |
| 182 |
| 183 /* Calculate sqrt(En * target_energy / excitation energy) */ |
| 184 targetEnergy = WebRtcSpl_Sqrt(dec_used_energy_); |
| 185 |
| 186 En = (int16_t) WebRtcSpl_Sqrt(En) << 6; |
| 187 En = (En * 3) >> 1; /* 1.5 estimates sqrt(2). */ |
| 188 dec_used_scale_factor_ = (int16_t)((En * targetEnergy) >> 12); |
| 189 |
| 190 /* Generate excitation. */ |
| 191 /* Excitation energy per sample is 2.^24 - Q13 N(0,1). */ |
| 192 for (size_t i = 0; i < num_samples; i++) { |
| 193 excitation[i] = WebRtcSpl_RandN(&dec_seed_) >> 1; |
| 194 } |
| 195 |
| 196 /* Scale to correct energy. */ |
| 197 WebRtcSpl_ScaleVector(excitation, excitation, dec_used_scale_factor_, |
| 198 num_samples, 13); |
| 199 |
| 200 /* |lpPoly| - Coefficients in Q12. |
| 201 * |excitation| - Speech samples. |
| 202 * |nst->dec_filtstate| - State preservation. |
| 203 * |out_data| - Filtered speech samples. */ |
| 204 WebRtcSpl_FilterAR(lpPoly, WEBRTC_CNG_MAX_LPC_ORDER + 1, excitation, |
| 205 num_samples, dec_filtstate_, WEBRTC_CNG_MAX_LPC_ORDER, |
| 206 dec_filtstateLow_, WEBRTC_CNG_MAX_LPC_ORDER, |
| 207 out_data.data(), low, num_samples); |
| 208 |
| 209 return true; |
| 210 } |
| 211 |
| 212 ComfortNoiseEncoder::ComfortNoiseEncoder(int fs, int interval, int quality) |
| 213 : enc_nrOfCoefs_(quality), |
| 214 enc_sampfreq_(fs), |
| 215 enc_interval_(interval), |
| 216 enc_msSinceSid_(0), |
| 217 enc_Energy_(0), |
| 218 enc_reflCoefs_{0}, |
| 219 enc_corrVector_{0}, |
| 220 enc_seed_(7777) /* For debugging only. */ { |
| 221 RTC_CHECK(quality <= WEBRTC_CNG_MAX_LPC_ORDER && quality > 0); |
| 222 /* Needed to get the right function pointers in SPLIB. */ |
| 223 WebRtcSpl_Init(); |
| 224 } |
| 225 |
| 226 void ComfortNoiseEncoder::Reset(int fs, int interval, int quality) { |
| 227 RTC_CHECK(quality <= WEBRTC_CNG_MAX_LPC_ORDER && quality > 0); |
| 228 enc_nrOfCoefs_ = quality; |
| 229 enc_sampfreq_ = fs; |
| 230 enc_interval_ = interval; |
| 231 enc_msSinceSid_ = 0; |
| 232 enc_Energy_ = 0; |
| 233 for (auto& c : enc_reflCoefs_) |
| 234 c = 0; |
| 235 for (auto& c : enc_corrVector_) |
| 236 c = 0; |
| 237 enc_seed_ = 7777; /* For debugging only. */ |
| 238 } |
| 239 |
| 240 size_t ComfortNoiseEncoder::Encode(rtc::ArrayView<const int16_t> speech, |
| 241 bool force_sid, |
| 242 rtc::Buffer* output) { |
| 243 int16_t arCoefs[WEBRTC_CNG_MAX_LPC_ORDER + 1]; |
| 244 int32_t corrVector[WEBRTC_CNG_MAX_LPC_ORDER + 1]; |
| 245 int16_t refCs[WEBRTC_CNG_MAX_LPC_ORDER + 1]; |
| 246 int16_t hanningW[kCngMaxOutsizeOrder]; |
| 247 int16_t ReflBeta = 19661; /* 0.6 in q15. */ |
| 248 int16_t ReflBetaComp = 13107; /* 0.4 in q15. */ |
| 249 int32_t outEnergy; |
| 250 int outShifts; |
| 251 size_t i; |
| 252 int stab; |
| 253 int acorrScale; |
| 254 size_t index; |
| 255 size_t ind, factor; |
| 256 int32_t* bptr; |
| 257 int32_t blo, bhi; |
| 258 int16_t negate; |
| 259 const int16_t* aptr; |
| 260 int16_t speechBuf[kCngMaxOutsizeOrder]; |
| 261 |
| 262 const size_t num_samples = speech.size(); |
| 263 RTC_CHECK_LE(num_samples, static_cast<size_t>(kCngMaxOutsizeOrder)); |
| 264 |
| 265 for (i = 0; i < num_samples; i++) { |
| 266 speechBuf[i] = speech[i]; |
| 267 } |
| 268 |
| 269 factor = num_samples; |
| 270 |
| 271 /* Calculate energy and a coefficients. */ |
| 272 outEnergy = WebRtcSpl_Energy(speechBuf, num_samples, &outShifts); |
| 273 while (outShifts > 0) { |
| 274 /* We can only do 5 shifts without destroying accuracy in |
| 275 * division factor. */ |
| 276 if (outShifts > 5) { |
| 277 outEnergy <<= (outShifts - 5); |
| 278 outShifts = 5; |
| 279 } else { |
| 280 factor /= 2; |
| 281 outShifts--; |
| 282 } |
| 283 } |
| 284 outEnergy = WebRtcSpl_DivW32W16(outEnergy, (int16_t)factor); |
| 285 |
| 286 if (outEnergy > 1) { |
| 287 /* Create Hanning Window. */ |
| 288 WebRtcSpl_GetHanningWindow(hanningW, num_samples / 2); |
| 289 for (i = 0; i < (num_samples / 2); i++) |
| 290 hanningW[num_samples - i - 1] = hanningW[i]; |
| 291 |
| 292 WebRtcSpl_ElementwiseVectorMult(speechBuf, hanningW, speechBuf, num_samples, |
| 293 14); |
| 294 |
| 295 WebRtcSpl_AutoCorrelation(speechBuf, num_samples, enc_nrOfCoefs_, |
| 296 corrVector, &acorrScale); |
| 297 |
| 298 if (*corrVector == 0) |
| 299 *corrVector = WEBRTC_SPL_WORD16_MAX; |
| 300 |
| 301 /* Adds the bandwidth expansion. */ |
| 302 aptr = WebRtcCng_kCorrWindow; |
| 303 bptr = corrVector; |
| 304 |
| 305 /* (zzz) lpc16_1 = 17+1+820+2+2 = 842 (ordo2=700). */ |
| 306 for (ind = 0; ind < enc_nrOfCoefs_; ind++) { |
| 307 /* The below code multiplies the 16 b corrWindow values (Q15) with |
| 308 * the 32 b corrvector (Q0) and shifts the result down 15 steps. */ |
| 309 negate = *bptr < 0; |
| 310 if (negate) |
| 311 *bptr = -*bptr; |
| 312 |
| 313 blo = (int32_t) * aptr * (*bptr & 0xffff); |
| 314 bhi = ((blo >> 16) & 0xffff) |
| 315 + ((int32_t)(*aptr++) * ((*bptr >> 16) & 0xffff)); |
| 316 blo = (blo & 0xffff) | ((bhi & 0xffff) << 16); |
| 317 |
| 318 *bptr = (((bhi >> 16) & 0x7fff) << 17) | ((uint32_t) blo >> 15); |
| 319 if (negate) |
| 320 *bptr = -*bptr; |
| 321 bptr++; |
| 322 } |
| 323 /* End of bandwidth expansion. */ |
| 324 |
| 325 stab = WebRtcSpl_LevinsonDurbin(corrVector, arCoefs, refCs, |
| 326 enc_nrOfCoefs_); |
| 327 |
| 328 if (!stab) { |
| 329 /* Disregard from this frame */ |
| 330 return 0; |
| 331 } |
| 332 |
| 333 } else { |
| 334 for (i = 0; i < enc_nrOfCoefs_; i++) |
| 335 refCs[i] = 0; |
| 336 } |
| 337 |
| 338 if (force_sid) { |
| 339 /* Read instantaneous values instead of averaged. */ |
| 340 for (i = 0; i < enc_nrOfCoefs_; i++) |
| 341 enc_reflCoefs_[i] = refCs[i]; |
| 342 enc_Energy_ = outEnergy; |
| 343 } else { |
| 344 /* Average history with new values. */ |
| 345 for (i = 0; i < enc_nrOfCoefs_; i++) { |
| 346 enc_reflCoefs_[i] = (int16_t) WEBRTC_SPL_MUL_16_16_RSFT( |
| 347 enc_reflCoefs_[i], ReflBeta, 15); |
| 348 enc_reflCoefs_[i] += |
| 349 (int16_t) WEBRTC_SPL_MUL_16_16_RSFT(refCs[i], ReflBetaComp, 15); |
| 350 } |
| 351 enc_Energy_ = |
| 352 (outEnergy >> 2) + (enc_Energy_ >> 1) + (enc_Energy_ >> 2); |
| 353 } |
| 354 |
| 355 if (enc_Energy_ < 1) { |
| 356 enc_Energy_ = 1; |
| 357 } |
| 358 |
| 359 if ((enc_msSinceSid_ > (enc_interval_ - 1)) || force_sid) { |
| 360 /* Search for best dbov value. */ |
| 361 index = 0; |
| 362 for (i = 1; i < 93; i++) { |
| 363 /* Always round downwards. */ |
| 364 if ((enc_Energy_ - WebRtcCng_kDbov[i]) > 0) { |
| 365 index = i; |
| 366 break; |
| 367 } |
| 368 } |
| 369 if ((i == 93) && (index == 0)) |
| 370 index = 94; |
| 371 |
| 372 const size_t output_coefs = enc_nrOfCoefs_ + 1; |
| 373 output->AppendData(output_coefs, [&] (rtc::ArrayView<uint8_t> output) { |
| 374 output[0] = (uint8_t)index; |
| 375 |
| 376 /* Quantize coefficients with tweak for WebRtc implementation of |
| 377 * RFC3389. */ |
| 378 if (enc_nrOfCoefs_ == WEBRTC_CNG_MAX_LPC_ORDER) { |
| 379 for (i = 0; i < enc_nrOfCoefs_; i++) { |
| 380 /* Q15 to Q7 with rounding. */ |
| 381 output[i + 1] = ((enc_reflCoefs_[i] + 128) >> 8); |
| 382 } |
| 383 } else { |
| 384 for (i = 0; i < enc_nrOfCoefs_; i++) { |
| 385 /* Q15 to Q7 with rounding. */ |
| 386 output[i + 1] = (127 + ((enc_reflCoefs_[i] + 128) >> 8)); |
| 387 } |
| 388 } |
| 389 |
| 390 return output_coefs; |
| 391 }); |
| 392 |
| 393 enc_msSinceSid_ = |
| 394 static_cast<int16_t>((1000 * num_samples) / enc_sampfreq_); |
| 395 return output_coefs; |
| 396 } else { |
| 397 enc_msSinceSid_ += |
| 398 static_cast<int16_t>((1000 * num_samples) / enc_sampfreq_); |
| 399 return 0; |
| 400 } |
| 401 } |
| 402 |
| 403 namespace { |
| 404 /* Values in |k| are Q15, and |a| Q12. */ |
| 405 void WebRtcCng_K2a16(int16_t* k, int useOrder, int16_t* a) { |
| 406 int16_t any[WEBRTC_SPL_MAX_LPC_ORDER + 1]; |
| 407 int16_t* aptr; |
| 408 int16_t* aptr2; |
| 409 int16_t* anyptr; |
| 410 const int16_t* kptr; |
| 411 int m, i; |
| 412 |
| 413 kptr = k; |
| 414 *a = 4096; /* i.e., (Word16_MAX >> 3) + 1 */ |
| 415 *any = *a; |
| 416 a[1] = (*k + 4) >> 3; |
| 417 for (m = 1; m < useOrder; m++) { |
| 418 kptr++; |
| 419 aptr = a; |
| 420 aptr++; |
| 421 aptr2 = &a[m]; |
| 422 anyptr = any; |
| 423 anyptr++; |
| 424 |
| 425 any[m + 1] = (*kptr + 4) >> 3; |
| 426 for (i = 0; i < m; i++) { |
| 427 *anyptr++ = |
| 428 (*aptr++) + |
| 429 (int16_t)((((int32_t)(*aptr2--) * (int32_t)*kptr) + 16384) >> 15); |
| 430 } |
| 431 |
| 432 aptr = a; |
| 433 anyptr = any; |
| 434 for (i = 0; i < (m + 2); i++) { |
| 435 *aptr++ = *anyptr++; |
| 436 } |
| 437 } |
| 438 } |
| 439 |
| 440 } // namespace |
| 441 |
| 442 } // namespace webrtc |
OLD | NEW |