From 381d69840ad3af2259f0b7ef49236f9ee9c76b76 Mon Sep 17 00:00:00 2001 From: Jean-Michel Trivi Date: Sun, 9 Sep 2012 11:48:51 -0700 Subject: Bug fixes in AAC decoder, AAC encoder, FDK library Latest code drop from Fraunhofer: * AAC-Decoder - Expanded AAC-LD/ELD decoder TNS max band tables to avoid wrong data access for sampling rates <22kHz and >48kHz. Modified file(s): libAACdec\src\aacdec_tns.cpp libAACdec\src\aac_rom.h libAACdec\src\aac_rom.cpp - Fixed ELD synthesis QMF filterbank scaling for downsampled SBR. Modified file(s): libFDK\src\qmf.cpp * AAC-Encoder - Fixed bit rate limiting for lower limit as introduced in the Delivery 2012-05-11. Modified file(s): libAACenc\include\aacenc_lib.h libAACenc\src\aacenc.h libAACenc\src\aacenc.cpp libSBRenc\src\sbr_rom.h libSBRenc\src\sbr_rom.cpp - Allow negative prediction gain as meaning that there is no coding gain. Make use of scaleValueSaturate in gauss window calculation. Modified file(s): libAACenc\src\aacenc_tns.cpp - Added energy saturation to prevent overflow in short blocks. Modified file(s): libAACenc\src\grp_data.cpp - Perform scalefactor adaption in case quantized lines are out of valid range. Modified file(s): libAACenc\src\quantize.cpp - Interrupt quantization loop when no spectral data is available. Adjustments in bitreservoir adaption. Modified file(s): libAACenc\src\qc_main.cpp - Indroduced dynamic scaling in none missing harmonic energy lowering compensation. Modified file(s): libSBRenc\src\env_est.cpp * FDK-Library - Fixed saturation for negative values in scaleValueSaturate. Modified file(s): libFDK\include\scale.h Change-Id: If830ea65caef6b5554281e4b7a77a8b2e08825ce --- libAACenc/include/aacenc_lib.h | 95 +----------------------------------------- libAACenc/src/aacenc.cpp | 11 ++--- libAACenc/src/aacenc.h | 1 - libAACenc/src/aacenc_lib.cpp | 2 +- libAACenc/src/aacenc_tns.cpp | 23 +++++----- libAACenc/src/grp_data.cpp | 12 ++++-- libAACenc/src/qc_main.cpp | 54 ++++++++++++++---------- libAACenc/src/quantize.cpp | 32 +++++++------- 8 files changed, 79 insertions(+), 151 deletions(-) (limited to 'libAACenc') diff --git a/libAACenc/include/aacenc_lib.h b/libAACenc/include/aacenc_lib.h index 4635119..862dcb5 100644 --- a/libAACenc/include/aacenc_lib.h +++ b/libAACenc/include/aacenc_lib.h @@ -424,102 +424,11 @@ For HE-AAC and HE-AAC v2 the lowest possible audio input sampling frequency is 1 AAC-LC core encoder operates in dual rate mode at its lowest possible sampling frequency, which is 8 kHz. HE-AAC v2 requires stereo input audio data. -The following table lists the supported bitrates for AAC-LC, HE-AAC and HE-AAC v2 encoding depending -on input sampling frequency ("Hz") and number of input channels ("chan"). The minimum and maximum -allowed bitrate ("BR Min", "BR Max") is given in bits per second. -In case the desired combination of bitrate and sampling frequency is not available ("NA") for HE-AAC or -HE-AAC v2 then the encoder will automatically switch to AAC-LC and give a command line warning. Please note that in HE-AAC or HE-AAC v2 mode the encoder supports much higher bitrates than are appropriate for HE-AAC or HE-AAC v2. For example, at a bitrate of more than 64 kbit/s for a stereo audio signal at 44.1 kHz it usually makes sense to use AAC-LC, which will produce better audio quality at that bitrate than HE-AAC or HE-AAC v2. - -\verbatim - Config AAC-LC HE-AAC (SBR) HE-AACv2 (SBR+PS) - - Hz chan BR Min BR Max BR Min BR Max BR Min BR Max - -8000 1 758 48000 NA NA NA NA -11025 1 1045 66150 NA NA NA NA -12000 1 1137 72000 NA NA NA NA -16000 1 1516 96000 8000 48000 NA NA -22050 1 2089 132300 8000 64000 NA NA -24000 1 2274 144000 8000 64000 NA NA -32000 1 3032 192000 8000 64000 NA NA -44100 1 4178 264576 8000 64000 NA NA -48000 1 4547 288000 12000 64000 NA NA -64000 1 6063 384000 24000 160000 NA NA -88200 1 8355 529200 24000 160000 NA NA -96000 1 9094 576000 24000 160000 NA NA ------------------------------------------------------------------------------------ -8000 2 1071 96000 NA NA NA NA -11025 2 1476 132300 NA NA NA NA -12000 2 1606 144000 NA NA NA NA -16000 2 2141 192000 16000 96000 8000 48000 -22050 2 2951 264600 16000 128000 8000 64000 -24000 2 3211 288000 16000 128000 8000 64000 -32000 2 4282 384000 16000 128000 8000 64000 -44100 2 5900 529152 16000 128000 8000 64000 -48000 2 6422 576000 16000 128000 12000 64000 -64000 2 8563 768000 32000 256000 24000 160000 -88200 2 11801 1058400 32000 256000 24000 160000 -96000 2 12844 1152000 32000 256000 24000 160000 ------------------------------------------------------------------------------------ -8000 3 1383 144000 NA NA NA NA -11025 3 1906 198450 NA NA NA NA -12000 3 2075 216000 NA NA NA NA -16000 3 2766 288000 26667 120000 NA NA -22050 3 3812 396900 26667 160000 NA NA -24000 3 4149 432000 26667 160000 NA NA -32000 3 5532 576000 26667 160000 NA NA -44100 3 7623 793728 26667 160000 NA NA -48000 3 8297 864000 29996 160000 NA NA -64000 3 11063 1152000 59996 400000 NA NA -88200 3 15246 1587600 59996 400000 NA NA -96000 3 16594 1728000 59996 400000 NA NA ------------------------------------------------------------------------------------ -8000 4 1696 192000 NA NA NA NA -11025 4 2337 264600 NA NA NA NA -12000 4 2543 288000 NA NA NA NA -16000 4 3391 384000 40000 160000 NA NA -22050 4 4673 529200 40000 213330 NA NA -24000 4 5086 576000 40000 213330 NA NA -32000 4 6782 768000 40000 213330 NA NA -44100 4 9345 1058304 40000 213330 NA NA -48000 4 10172 1152000 40000 213330 NA NA -64000 4 13563 1536000 80000 533330 NA NA -88200 4 18691 2116800 80000 533330 NA NA -96000 4 20344 2304000 80000 533330 NA NA ------------------------------------------------------------------------------------ -8000 5 2008 240000 NA NA NA NA -11025 5 2768 330750 NA NA NA NA -12000 5 3012 360000 NA NA NA NA -16000 5 4016 480000 43244 184612 NA NA -22050 5 5535 661500 43244 246152 NA NA -24000 5 6024 720000 43244 246152 NA NA -32000 5 8032 960000 43244 246152 NA NA -44100 5 11068 1322880 43244 246152 NA NA -48000 5 12047 1440000 46140 246152 NA NA -64000 5 16063 1920000 92296 615384 NA NA -88200 5 22137 2646000 92296 615384 NA NA -96000 5 24094 2880000 92296 615384 NA NA ------------------------------------------------------------------------------------ -8000 5.1 2321 240000 NA NA NA NA -11025 5.1 3198 330750 NA NA NA NA -12000 5.1 3481 360000 NA NA NA NA -16000 5.1 4641 480000 45715 199990 NA NA -22050 5.1 6396 661500 45715 266658 NA NA -24000 5.1 6961 720000 45715 266658 NA NA -32000 5.1 9282 960000 45715 266658 NA NA -44100 5.1 12790 1322880 45715 266658 NA NA -48000 5.1 13922 1440000 49982 266658 NA NA -64000 5.1 18563 1920000 99982 666658 NA NA -88200 5.1 25582 2646000 99982 666658 NA NA -96000 5.1 27844 2880000 99982 666658 NA NA - -\endverbatim \n - \section reommendedConfig Recommended Sampling Rate and Bitrate Combinations The following table provides an overview of recommended encoder configuration parameters @@ -956,8 +865,8 @@ typedef enum AACENC_GRANULE_LENGTH = 0x0105, /*!< Core encoder (AAC) audio frame length in samples: - 1024: Default configuration. - - 512: Optional length in LD/ELD configuration. - - 480: Default LD/ELD configuration. */ + - 512: Default LD/ELD configuration. + - 480: Optional length in LD/ELD configuration. */ AACENC_CHANNELMODE = 0x0106, /*!< Set explicit channel mode. Channel mode must match with number of input channels. - 1-6: MPEG channel modes supported, see ::CHANNEL_MODE in FDK_audio.h. */ diff --git a/libAACenc/src/aacenc.cpp b/libAACenc/src/aacenc.cpp index 85083cd..d3f36aa 100644 --- a/libAACenc/src/aacenc.cpp +++ b/libAACenc/src/aacenc.cpp @@ -149,7 +149,7 @@ INT FDKaacEnc_LimitBitrate( transportBits = 208; } - bitRate = FDKmax(bitRate, ((((40 * nChannels) + transportBits + frameLength) * (coreSamplingRate)) / frameLength) ); + bitRate = FDKmax(bitRate, ((((40 * nChannels) + transportBits) * (coreSamplingRate)) / frameLength) ); FDK_ASSERT(bitRate >= 0); bitRate = FDKmin(bitRate, ((nChannelsEff * MIN_BUFSIZE_PER_EFF_CHAN)*(coreSamplingRate>>shift)) / (frameLength>>shift)) ; @@ -280,7 +280,7 @@ void FDKaacEnc_AacInitDefaultConfig(AACENC_CONFIG *config) config->useTns = TNS_ENABLE_MASK; /* tns enabled completly */ config->usePns = 1; /* depending on channelBitrate this might be set to 0 later */ config->useIS = 1; /* Intensity Stereo Configuration */ - config->framelength = DEFAULT_FRAMELENGTH; /* used frame size */ + config->framelength = -1; /* Framesize not configured */ config->syntaxFlags = 0; /* default syntax with no specialities */ config->epConfig = -1; /* no ER syntax -> no additional error protection */ config->nSubFrames = 1; /* default, no sub frames */ @@ -451,11 +451,8 @@ AAC_ENCODER_ERROR FDKaacEnc_Initialize(HANDLE_AAC_ENC hAacEnc, switch (config->framelength) { case 1024: - if ( config->audioObjectType != AOT_AAC_LC - && config->audioObjectType != AOT_SBR - && config->audioObjectType != AOT_PS - && config->audioObjectType != AOT_ER_AAC_LC - && config->audioObjectType != AOT_AAC_SCAL ) + if ( config->audioObjectType == AOT_ER_AAC_LD + || config->audioObjectType == AOT_ER_AAC_ELD ) { return AAC_ENC_INVALID_FRAME_LENGTH; } diff --git a/libAACenc/src/aacenc.h b/libAACenc/src/aacenc.h index 8242248..942f686 100644 --- a/libAACenc/src/aacenc.h +++ b/libAACenc/src/aacenc.h @@ -153,7 +153,6 @@ typedef enum { /*-------------------------- defines --------------------------------------*/ #define ANC_DATA_BUFFERSIZE 1024 /* ancBuffer size */ -#define DEFAULT_FRAMELENGTH 1024 /* size of AAC core frame in (new) PCM samples */ #define MAX_TOTAL_EXT_PAYLOADS (((6) * (1)) + (2+2)) diff --git a/libAACenc/src/aacenc_lib.cpp b/libAACenc/src/aacenc_lib.cpp index cbb0e2a..a245358 100644 --- a/libAACenc/src/aacenc_lib.cpp +++ b/libAACenc/src/aacenc_lib.cpp @@ -98,7 +98,7 @@ amm-info@iis.fraunhofer.de /* Encoder library info */ #define AACENCODER_LIB_VL0 3 #define AACENCODER_LIB_VL1 3 -#define AACENCODER_LIB_VL2 1 +#define AACENCODER_LIB_VL2 2 #define AACENCODER_LIB_TITLE "AAC Encoder" #define AACENCODER_LIB_BUILD_DATE __DATE__ #define AACENCODER_LIB_BUILD_TIME __TIME__ diff --git a/libAACenc/src/aacenc_tns.cpp b/libAACenc/src/aacenc_tns.cpp index 933e4e7..00b2bca 100644 --- a/libAACenc/src/aacenc_tns.cpp +++ b/libAACenc/src/aacenc_tns.cpp @@ -1067,11 +1067,11 @@ static void FDKaacEnc_CalcGaussWindow( const INT timeResolution_e ) { - #define PI_SCALE (2) - #define PI_FIX FL2FXCONST_DBL(3.1416f/(float)(1<=(FIXP_DBL)MAXVAL_DBL-b) ? (FIXP_DBL)MAXVAL_DBL : (a + b) ); +} + void FDKaacEnc_groupShortData(FIXP_DBL *mdctSpectrum, /* in-out */ SFB_THRESHOLD *sfbThreshold, /* in-out */ @@ -177,7 +181,7 @@ FDKaacEnc_groupShortData(FIXP_DBL *mdctSpectrum, /* in-out FIXP_DBL thresh = sfbThreshold->Short[wnd][sfb]; for (j=1; jShort[wnd+j][sfb]; + thresh = nrgAddSaturate(thresh, sfbThreshold->Short[wnd+j][sfb]); } sfbThreshold->Long[i++] = thresh; } @@ -195,7 +199,7 @@ FDKaacEnc_groupShortData(FIXP_DBL *mdctSpectrum, /* in-out FIXP_DBL energy = sfbEnergy->Short[wnd][sfb]; for (j=1; jShort[wnd+j][sfb]; + energy = nrgAddSaturate(energy, sfbEnergy->Short[wnd+j][sfb]); } sfbEnergy->Long[i++] = energy; } @@ -213,7 +217,7 @@ FDKaacEnc_groupShortData(FIXP_DBL *mdctSpectrum, /* in-out FIXP_DBL energy = sfbEnergyMS->Short[wnd][sfb]; for (j=1; jShort[wnd+j][sfb]; + energy = nrgAddSaturate(energy, sfbEnergyMS->Short[wnd+j][sfb]); } sfbEnergyMS->Long[i++] = energy; } @@ -231,7 +235,7 @@ FDKaacEnc_groupShortData(FIXP_DBL *mdctSpectrum, /* in-out FIXP_DBL energy = sfbSpreadEnergy->Short[wnd][sfb]; for (j=1; jShort[wnd+j][sfb]; + energy = nrgAddSaturate(energy, sfbSpreadEnergy->Short[wnd+j][sfb]); } sfbSpreadEnergy->Long[i++] = energy; } diff --git a/libAACenc/src/qc_main.cpp b/libAACenc/src/qc_main.cpp index 749398a..d7e76c7 100644 --- a/libAACenc/src/qc_main.cpp +++ b/libAACenc/src/qc_main.cpp @@ -797,7 +797,7 @@ AAC_ENCODER_ERROR FDKaacEnc_QCMain(QC_STATE* RESTRICT hQC, { int i, c; AAC_ENCODER_ERROR ErrorStatus = AAC_ENC_OK; - INT avgTotalDynBits = 0; /* maximal allowd dynamic bits for all frames */ + INT avgTotalDynBits = 0; /* maximal allowed dynamic bits for all frames */ INT totalAvailableBits = 0; INT nSubFrames = 1; @@ -1092,7 +1092,7 @@ AAC_ENCODER_ERROR FDKaacEnc_QCMain(QC_STATE* RESTRICT hQC, int sumBitsConsumedTotal = FDKaacEnc_getTotalConsumedBits(qcOut, qcElement, cm, hQC->globHdrBits, nSubFrames); /* in all frames are valid dynamic bits */ - if (sumBitsConsumedTotal < totalAvailableBits && (decreaseBitConsumption==1) && checkMinFrameBitsDemand(qcOut,hQC->minBitsPerFrame,nSubFrames) + if ( ((sumBitsConsumedTotal < totalAvailableBits) || qcOut[c]->usedDynBits==0) && (decreaseBitConsumption==1) && checkMinFrameBitsDemand(qcOut,hQC->minBitsPerFrame,nSubFrames) /*()*/ ) { quantizationDone = 1; /* exit bit adjustment */ @@ -1365,42 +1365,54 @@ AAC_ENCODER_ERROR FDKaacEnc_FinalizeBitConsumption(CHANNEL_MAPPING *cm, QC_OUT_EXTENSION fillExtPayload; INT totFillBits, alignBits; - { - int exactTpBits; - int max_iter = 3; + /* Get total consumed bits in AU */ + qcOut->totalBits = qcOut->staticBits + qcOut->usedDynBits + qcOut->totFillBits + + qcOut->elementExtBits + qcOut->globalExtBits; - /* Get total consumed bits in AU */ - qcOut->totalBits = qcOut->staticBits + qcOut->usedDynBits + qcOut->totFillBits + - qcOut->elementExtBits + qcOut->globalExtBits; + if (qcKernel->bitrateMode==QCDATA_BR_MODE_CBR) { /* Now we can get the exact transport bit amount, and hopefully it is equal to the estimated value */ - exactTpBits = transportEnc_GetStaticBits(hTpEnc, qcOut->totalBits); + INT exactTpBits = transportEnc_GetStaticBits(hTpEnc, qcOut->totalBits); if (exactTpBits != qcKernel->globHdrBits) { INT diffFillBits = 0; + /* How many bits can be taken by bitreservoir */ + const INT bitresSpace = qcKernel->bitResTotMax - (qcKernel->bitResTot + (qcOut->grantedDynBits - (qcOut->usedDynBits + qcOut->totFillBits) ) ); + /* Number of bits which can be moved to bitreservoir. */ - INT bitsToBitres = qcKernel->globHdrBits - exactTpBits; + const INT bitsToBitres = qcKernel->globHdrBits - exactTpBits; + FDK_ASSERT(bitsToBitres>=0); /* is always positive */ - if (bitsToBitres>0) { - /* if bitreservoir can not take all bits, move ramaining bits to fillbits */ - diffFillBits = FDKmax(0, bitsToBitres - (qcKernel->bitResTotMax-qcKernel->bitResTot)); - } - else if (bitsToBitres<0) { - /* if bits mus be taken from bitreservoir, reduce fillbits first. */ - diffFillBits = (FDKmax(FDKmax(bitsToBitres, -qcKernel->bitResTot), -qcOut->totFillBits)); - } + /* If bitreservoir can not take all bits, move ramaining bits to fillbits */ + diffFillBits = FDKmax(0, bitsToBitres - bitresSpace); + + /* Assure previous alignment */ + diffFillBits = (diffFillBits+7)&~7; - diffFillBits = (diffFillBits+7)&~7; /* assure previous alignment */ + /* Move as many bits as possible to bitreservoir */ + qcKernel->bitResTot += (bitsToBitres-diffFillBits); + /* Write remaing bits as fill bits */ qcOut->totFillBits += diffFillBits; qcOut->totalBits += diffFillBits; qcOut->grantedDynBits += diffFillBits; - /* new header bits */ + /* Get new header bits */ qcKernel->globHdrBits = transportEnc_GetStaticBits(hTpEnc, qcOut->totalBits); + + if (qcKernel->globHdrBits != exactTpBits) { + /* In previous step, fill bits and corresponding total bits were changed when bitreservoir was completely filled. + Now we can take the too much taken bits caused by header overhead from bitreservoir. + */ + qcKernel->bitResTot -= (qcKernel->globHdrBits - exactTpBits); + } } - } + + } /* MODE_CBR */ + + /* Update exact number of consumed header bits. */ + qcKernel->globHdrBits = transportEnc_GetStaticBits(hTpEnc, qcOut->totalBits); /* Save total fill bits and distribut to alignment and fill bits */ totFillBits = qcOut->totFillBits; diff --git a/libAACenc/src/quantize.cpp b/libAACenc/src/quantize.cpp index dc85a6d..a1698a8 100644 --- a/libAACenc/src/quantize.cpp +++ b/libAACenc/src/quantize.cpp @@ -127,10 +127,7 @@ static void FDKaacEnc_quantizeLines(INT gain, accu = fMultDiv2(FDKaacEnc_mTab_3_4[tabIndex],FDKaacEnc_quantTableE[totalShift&3]); totalShift = (16-4)-(3*(totalShift>>2)); FDK_ASSERT(totalShift >=0); /* MAX_QUANT_VIOLATION */ - if (totalShift < 32) - accu>>=totalShift; - else - accu = 0; + accu>>=totalShift; quaSpectrum[line] = (SHORT)(-((LONG)(k + accu) >> (DFRACT_BITS-1-16))); } else if(accu > FL2FXCONST_DBL(0.0f)) @@ -143,10 +140,7 @@ static void FDKaacEnc_quantizeLines(INT gain, accu = fMultDiv2(FDKaacEnc_mTab_3_4[tabIndex],FDKaacEnc_quantTableE[totalShift&3]); totalShift = (16-4)-(3*(totalShift>>2)); FDK_ASSERT(totalShift >=0); /* MAX_QUANT_VIOLATION */ - if (totalShift < 32) - accu>>=totalShift; - else - accu = 0; + accu>>=totalShift; quaSpectrum[line] = (SHORT)((LONG)(k + accu) >> (DFRACT_BITS-1-16)); } else @@ -319,6 +313,9 @@ FIXP_DBL FDKaacEnc_calcSfbDist(FIXP_DBL *mdctSpectrum, &mdctSpectrum[i], &quantSpectrum[i]); + if (fAbs(quantSpectrum[i])>MAX_QUANT) { + return FL2FXCONST_DBL(0.0f); + } /* inverse quantization */ FDKaacEnc_invQuantizeLines(gain,1,&quantSpectrum[i],&invQuantSpec); @@ -361,15 +358,22 @@ void FDKaacEnc_calcSfbQuantEnergyAndDist(FIXP_DBL *mdctSpectrum, FIXP_DBL invQuantSpec; FIXP_DBL diff; - *en = FL2FXCONST_DBL(0.0f); - *dist = FL2FXCONST_DBL(0.0f); + FIXP_DBL energy = FL2FXCONST_DBL(0.0f); + FIXP_DBL distortion = FL2FXCONST_DBL(0.0f); for (i=0; iMAX_QUANT) { + *en = FL2FXCONST_DBL(0.0f); + *dist = FL2FXCONST_DBL(0.0f); + return; + } + /* inverse quantization */ FDKaacEnc_invQuantizeLines(gain,1,&quantSpectrum[i],&invQuantSpec); /* energy */ - *en += fPow2(invQuantSpec); + energy += fPow2(invQuantSpec); /* dist */ diff = fixp_abs(fixp_abs(invQuantSpec) - fixp_abs(mdctSpectrum[i]>>1)); @@ -382,10 +386,10 @@ void FDKaacEnc_calcSfbQuantEnergyAndDist(FIXP_DBL *mdctSpectrum, diff = scaleValue(diff, -scale); - *dist += diff; + distortion += diff; } - *en = CalcLdData(*en)+FL2FXCONST_DBL(0.03125f); - *dist = CalcLdData(*dist); + *en = CalcLdData(energy)+FL2FXCONST_DBL(0.03125f); + *dist = CalcLdData(distortion); } -- cgit v1.2.3