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 * structs.h 13 * 14 * This header file contains all the structs used in the ISAC codec 15 * 16 */ 17 18 #ifndef WEBRTC_MODULES_AUDIO_CODING_CODECS_ISAC_MAIN_SOURCE_STRUCTS_H_ 19 #define WEBRTC_MODULES_AUDIO_CODING_CODECS_ISAC_MAIN_SOURCE_STRUCTS_H_ 20 21 #include "webrtc/modules/audio_coding/codecs/isac/main/interface/isac.h" 22 #include "webrtc/modules/audio_coding/codecs/isac/main/source/settings.h" 23 #include "webrtc/typedefs.h" 24 25 typedef struct Bitstreamstruct { 26 27 uint8_t stream[STREAM_SIZE_MAX]; 28 uint32_t W_upper; 29 uint32_t streamval; 30 uint32_t stream_index; 31 32 } Bitstr; 33 34 typedef struct { 35 36 double DataBufferLo[WINLEN]; 37 double DataBufferHi[WINLEN]; 38 39 double CorrBufLo[ORDERLO+1]; 40 double CorrBufHi[ORDERHI+1]; 41 42 float PreStateLoF[ORDERLO+1]; 43 float PreStateLoG[ORDERLO+1]; 44 float PreStateHiF[ORDERHI+1]; 45 float PreStateHiG[ORDERHI+1]; 46 float PostStateLoF[ORDERLO+1]; 47 float PostStateLoG[ORDERLO+1]; 48 float PostStateHiF[ORDERHI+1]; 49 float PostStateHiG[ORDERHI+1]; 50 51 double OldEnergy; 52 53 } MaskFiltstr; 54 55 56 typedef struct { 57 58 //state vectors for each of the two analysis filters 59 double INSTAT1[2*(QORDER-1)]; 60 double INSTAT2[2*(QORDER-1)]; 61 double INSTATLA1[2*(QORDER-1)]; 62 double INSTATLA2[2*(QORDER-1)]; 63 double INLABUF1[QLOOKAHEAD]; 64 double INLABUF2[QLOOKAHEAD]; 65 66 float INSTAT1_float[2*(QORDER-1)]; 67 float INSTAT2_float[2*(QORDER-1)]; 68 float INSTATLA1_float[2*(QORDER-1)]; 69 float INSTATLA2_float[2*(QORDER-1)]; 70 float INLABUF1_float[QLOOKAHEAD]; 71 float INLABUF2_float[QLOOKAHEAD]; 72 73 /* High pass filter */ 74 double HPstates[HPORDER]; 75 float HPstates_float[HPORDER]; 76 77 } PreFiltBankstr; 78 79 80 typedef struct { 81 82 //state vectors for each of the two analysis filters 83 double STATE_0_LOWER[2*POSTQORDER]; 84 double STATE_0_UPPER[2*POSTQORDER]; 85 86 /* High pass filter */ 87 double HPstates1[HPORDER]; 88 double HPstates2[HPORDER]; 89 90 float STATE_0_LOWER_float[2*POSTQORDER]; 91 float STATE_0_UPPER_float[2*POSTQORDER]; 92 93 float HPstates1_float[HPORDER]; 94 float HPstates2_float[HPORDER]; 95 96 } PostFiltBankstr; 97 98 typedef struct { 99 100 //data buffer for pitch filter 101 double ubuf[PITCH_BUFFSIZE]; 102 103 //low pass state vector 104 double ystate[PITCH_DAMPORDER]; 105 106 //old lag and gain 107 double oldlagp[1]; 108 double oldgainp[1]; 109 110 } PitchFiltstr; 111 112 typedef struct { 113 114 //data buffer 115 double buffer[PITCH_WLPCBUFLEN]; 116 117 //state vectors 118 double istate[PITCH_WLPCORDER]; 119 double weostate[PITCH_WLPCORDER]; 120 double whostate[PITCH_WLPCORDER]; 121 122 //LPC window -> should be a global array because constant 123 double window[PITCH_WLPCWINLEN]; 124 125 } WeightFiltstr; 126 127 typedef struct { 128 129 //for inital estimator 130 double dec_buffer[PITCH_CORR_LEN2 + PITCH_CORR_STEP2 + 131 PITCH_MAX_LAG/2 - PITCH_FRAME_LEN/2+2]; 132 double decimator_state[2*ALLPASSSECTIONS+1]; 133 double hp_state[2]; 134 135 double whitened_buf[QLOOKAHEAD]; 136 137 double inbuf[QLOOKAHEAD]; 138 139 PitchFiltstr PFstr_wght; 140 PitchFiltstr PFstr; 141 WeightFiltstr Wghtstr; 142 143 } PitchAnalysisStruct; 144 145 146 147 /* Have instance of struct together with other iSAC structs */ 148 typedef struct { 149 150 /* Previous frame length (in ms) */ 151 int32_t prev_frame_length; 152 153 /* Previous RTP timestamp from received 154 packet (in samples relative beginning) */ 155 int32_t prev_rec_rtp_number; 156 157 /* Send timestamp for previous packet (in ms using timeGetTime()) */ 158 uint32_t prev_rec_send_ts; 159 160 /* Arrival time for previous packet (in ms using timeGetTime()) */ 161 uint32_t prev_rec_arr_ts; 162 163 /* rate of previous packet, derived from RTP timestamps (in bits/s) */ 164 float prev_rec_rtp_rate; 165 166 /* Time sinse the last update of the BN estimate (in ms) */ 167 uint32_t last_update_ts; 168 169 /* Time sinse the last reduction (in ms) */ 170 uint32_t last_reduction_ts; 171 172 /* How many times the estimate was update in the beginning */ 173 int32_t count_tot_updates_rec; 174 175 /* The estimated bottle neck rate from there to here (in bits/s) */ 176 int32_t rec_bw; 177 float rec_bw_inv; 178 float rec_bw_avg; 179 float rec_bw_avg_Q; 180 181 /* The estimated mean absolute jitter value, 182 as seen on this side (in ms) */ 183 float rec_jitter; 184 float rec_jitter_short_term; 185 float rec_jitter_short_term_abs; 186 float rec_max_delay; 187 float rec_max_delay_avg_Q; 188 189 /* (assumed) bitrate for headers (bps) */ 190 float rec_header_rate; 191 192 /* The estimated bottle neck rate from here to there (in bits/s) */ 193 float send_bw_avg; 194 195 /* The estimated mean absolute jitter value, as seen on 196 the other siee (in ms) */ 197 float send_max_delay_avg; 198 199 // number of packets received since last update 200 int num_pkts_rec; 201 202 int num_consec_rec_pkts_over_30k; 203 204 // flag for marking that a high speed network has been 205 // detected downstream 206 int hsn_detect_rec; 207 208 int num_consec_snt_pkts_over_30k; 209 210 // flag for marking that a high speed network has 211 // been detected upstream 212 int hsn_detect_snd; 213 214 uint32_t start_wait_period; 215 216 int in_wait_period; 217 218 int change_to_WB; 219 220 uint32_t senderTimestamp; 221 uint32_t receiverTimestamp; 222 //enum IsacSamplingRate incomingStreamSampFreq; 223 uint16_t numConsecLatePkts; 224 float consecLatency; 225 int16_t inWaitLatePkts; 226 } BwEstimatorstr; 227 228 229 typedef struct { 230 231 /* boolean, flags if previous packet exceeded B.N. */ 232 int PrevExceed; 233 /* ms */ 234 int ExceedAgo; 235 /* packets left to send in current burst */ 236 int BurstCounter; 237 /* packets */ 238 int InitCounter; 239 /* ms remaining in buffer when next packet will be sent */ 240 double StillBuffered; 241 242 } RateModel; 243 244 245 typedef struct { 246 247 unsigned int SpaceAlloced; 248 unsigned int MaxPermAlloced; 249 double Tmp0[MAXFFTSIZE]; 250 double Tmp1[MAXFFTSIZE]; 251 double Tmp2[MAXFFTSIZE]; 252 double Tmp3[MAXFFTSIZE]; 253 int Perm[MAXFFTSIZE]; 254 int factor [NFACTOR]; 255 256 } FFTstr; 257 258 259 /* The following strutc is used to store data from encoding, to make it 260 fast and easy to construct a new bitstream with a different Bandwidth 261 estimate. All values (except framelength and minBytes) is double size to 262 handle 60 ms of data. 263 */ 264 typedef struct { 265 266 /* Used to keep track of if it is first or second part of 60 msec packet */ 267 int startIdx; 268 269 /* Frame length in samples */ 270 int16_t framelength; 271 272 /* Pitch Gain */ 273 int pitchGain_index[2]; 274 275 /* Pitch Lag */ 276 double meanGain[2]; 277 int pitchIndex[PITCH_SUBFRAMES*2]; 278 279 /* LPC */ 280 int LPCindex_s[108*2]; /* KLT_ORDER_SHAPE = 108 */ 281 int LPCindex_g[12*2]; /* KLT_ORDER_GAIN = 12 */ 282 double LPCcoeffs_lo[(ORDERLO+1)*SUBFRAMES*2]; 283 double LPCcoeffs_hi[(ORDERHI+1)*SUBFRAMES*2]; 284 285 /* Encode Spec */ 286 int16_t fre[FRAMESAMPLES]; 287 int16_t fim[FRAMESAMPLES]; 288 int16_t AvgPitchGain[2]; 289 290 /* Used in adaptive mode only */ 291 int minBytes; 292 293 } ISAC_SaveEncData_t; 294 295 296 typedef struct { 297 298 int indexLPCShape[UB_LPC_ORDER * UB16_LPC_VEC_PER_FRAME]; 299 double lpcGain[SUBFRAMES<<1]; 300 int lpcGainIndex[SUBFRAMES<<1]; 301 302 Bitstr bitStreamObj; 303 304 int16_t realFFT[FRAMESAMPLES_HALF]; 305 int16_t imagFFT[FRAMESAMPLES_HALF]; 306 } ISACUBSaveEncDataStruct; 307 308 309 310 typedef struct { 311 312 Bitstr bitstr_obj; 313 MaskFiltstr maskfiltstr_obj; 314 PreFiltBankstr prefiltbankstr_obj; 315 PitchFiltstr pitchfiltstr_obj; 316 PitchAnalysisStruct pitchanalysisstr_obj; 317 FFTstr fftstr_obj; 318 ISAC_SaveEncData_t SaveEnc_obj; 319 320 int buffer_index; 321 int16_t current_framesamples; 322 323 float data_buffer_float[FRAMESAMPLES_30ms]; 324 325 int frame_nb; 326 double bottleneck; 327 int16_t new_framelength; 328 double s2nr; 329 330 /* Maximum allowed number of bits for a 30 msec packet */ 331 int16_t payloadLimitBytes30; 332 /* Maximum allowed number of bits for a 30 msec packet */ 333 int16_t payloadLimitBytes60; 334 /* Maximum allowed number of bits for both 30 and 60 msec packet */ 335 int16_t maxPayloadBytes; 336 /* Maximum allowed rate in bytes per 30 msec packet */ 337 int16_t maxRateInBytes; 338 339 /*--- 340 If set to 1 iSAC will not addapt the frame-size, if used in 341 channel-adaptive mode. The initial value will be used for all rates. 342 ---*/ 343 int16_t enforceFrameSize; 344 345 /*----- 346 This records the BWE index the encoder injected into the bit-stream. 347 It will be used in RCU. The same BWE index of main payload will be in 348 the redundant payload. We can not retrive it from BWE because it is 349 a recursive procedure (WebRtcIsac_GetDownlinkBwJitIndexImpl) and has to be 350 called only once per each encode. 351 -----*/ 352 int16_t lastBWIdx; 353 } ISACLBEncStruct; 354 355 typedef struct { 356 357 Bitstr bitstr_obj; 358 MaskFiltstr maskfiltstr_obj; 359 PreFiltBankstr prefiltbankstr_obj; 360 FFTstr fftstr_obj; 361 ISACUBSaveEncDataStruct SaveEnc_obj; 362 363 int buffer_index; 364 float data_buffer_float[MAX_FRAMESAMPLES + 365 LB_TOTAL_DELAY_SAMPLES]; 366 double bottleneck; 367 /* Maximum allowed number of bits for a 30 msec packet */ 368 //int16_t payloadLimitBytes30; 369 /* Maximum allowed number of bits for both 30 and 60 msec packet */ 370 //int16_t maxPayloadBytes; 371 int16_t maxPayloadSizeBytes; 372 373 double lastLPCVec[UB_LPC_ORDER]; 374 int16_t numBytesUsed; 375 int16_t lastJitterInfo; 376 } ISACUBEncStruct; 377 378 379 380 typedef struct { 381 382 Bitstr bitstr_obj; 383 MaskFiltstr maskfiltstr_obj; 384 PostFiltBankstr postfiltbankstr_obj; 385 PitchFiltstr pitchfiltstr_obj; 386 FFTstr fftstr_obj; 387 388 } ISACLBDecStruct; 389 390 typedef struct { 391 392 Bitstr bitstr_obj; 393 MaskFiltstr maskfiltstr_obj; 394 PostFiltBankstr postfiltbankstr_obj; 395 FFTstr fftstr_obj; 396 397 } ISACUBDecStruct; 398 399 400 401 typedef struct { 402 403 ISACLBEncStruct ISACencLB_obj; 404 ISACLBDecStruct ISACdecLB_obj; 405 } ISACLBStruct; 406 407 408 typedef struct { 409 410 ISACUBEncStruct ISACencUB_obj; 411 ISACUBDecStruct ISACdecUB_obj; 412 } ISACUBStruct; 413 414 /* 415 This struct is used to take a snapshot of the entropy coder and LPC gains 416 right before encoding LPC gains. This allows us to go back to that state 417 if we like to limit the payload size. 418 */ 419 typedef struct { 420 /* 6 lower-band & 6 upper-band */ 421 double loFiltGain[SUBFRAMES]; 422 double hiFiltGain[SUBFRAMES]; 423 /* Upper boundary of interval W */ 424 uint32_t W_upper; 425 uint32_t streamval; 426 /* Index to the current position in bytestream */ 427 uint32_t stream_index; 428 uint8_t stream[3]; 429 } transcode_obj; 430 431 432 typedef struct { 433 // lower-band codec instance 434 ISACLBStruct instLB; 435 // upper-band codec instance 436 ISACUBStruct instUB; 437 438 // Bandwidth Estimator and model for the rate. 439 BwEstimatorstr bwestimator_obj; 440 RateModel rate_data_obj; 441 double MaxDelay; 442 443 /* 0 = adaptive; 1 = instantaneous */ 444 int16_t codingMode; 445 446 // overall bottleneck of the codec 447 int32_t bottleneck; 448 449 // QMF Filter state 450 int32_t analysisFBState1[FB_STATE_SIZE_WORD32]; 451 int32_t analysisFBState2[FB_STATE_SIZE_WORD32]; 452 int32_t synthesisFBState1[FB_STATE_SIZE_WORD32]; 453 int32_t synthesisFBState2[FB_STATE_SIZE_WORD32]; 454 455 // Error Code 456 int16_t errorCode; 457 458 // bandwidth of the encoded audio 8, 12 or 16 kHz 459 enum ISACBandwidth bandwidthKHz; 460 // Sampling rate of audio, encoder and decode, 8 or 16 kHz 461 enum IsacSamplingRate encoderSamplingRateKHz; 462 enum IsacSamplingRate decoderSamplingRateKHz; 463 // Flag to keep track of initializations, lower & upper-band 464 // encoder and decoder. 465 int16_t initFlag; 466 467 // Flag to to indicate signal bandwidth switch 468 int16_t resetFlag_8kHz; 469 470 // Maximum allowed rate, measured in Bytes per 30 ms. 471 int16_t maxRateBytesPer30Ms; 472 // Maximum allowed payload-size, measured in Bytes. 473 int16_t maxPayloadSizeBytes; 474 /* The expected sampling rate of the input signal. Valid values are 16000, 475 * 32000 and 48000. This is not the operation sampling rate of the codec. 476 * Input signals at 48 kHz are resampled to 32 kHz, then encoded. */ 477 uint16_t in_sample_rate_hz; 478 /* State for the input-resampler. It is only used for 48 kHz input signals. */ 479 int16_t state_in_resampler[SIZE_RESAMPLER_STATE]; 480 } ISACMainStruct; 481 482 #endif /* WEBRTC_MODULES_AUDIO_CODING_CODECS_ISAC_MAIN_SOURCE_STRUCTS_H_ */ 483
