fenix/exhale
1
0
Fork 0
mirror of https://gitlab.com/ecodis/exhale.git synced 2025-06-05 21:59:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

prepare band M/S

Browse Source
This commit is contained in:
Christian R. Helmrich
2020-04-19 00:00:03 +02:00
parent 9ece3e4f70
commit 1fd402f60c
5 changed files with 91 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
src/lib/bitStreamWriter.cpp
View File

@ -11,7 +11,12 @@
#include "exhaleLibPch.h" #include "exhaleLibPch.h"
#include "bitStreamWriter.h" #include "bitStreamWriter.h"
// static helper function // static helper functions
static inline int getPredCoefPrevGrp (const uint8_t aqIdxPrevGrp)
{
return int (aqIdxPrevGrp > 0 ? aqIdxPrevGrp & 31 : 16);
}
static uint32_t getDeltaCodeTimeFlag (const uint8_t* const alphaQCurr, const unsigned numWinGroups, const unsigned numSwbShort, static uint32_t getDeltaCodeTimeFlag (const uint8_t* const alphaQCurr, const unsigned numWinGroups, const unsigned numSwbShort,
const uint8_t* const alphaQPrev, const unsigned maxSfbSte, const EntropyCoder& entrCoder, const uint8_t* const alphaQPrev, const unsigned maxSfbSte, const EntropyCoder& entrCoder,
const bool complexCoef) const bool complexCoef)
@ -34,7 +39,7 @@ static uint32_t getDeltaCodeTimeFlag (const uint8_t* const alphaQCurr, const uns
int aqIdx = gCplxPredUsed[b] & 31; // range -15,...0,...,15 int aqIdx = gCplxPredUsed[b] & 31; // range -15,...0,...,15
bitCountFreq += entrCoder.indexGetBitCount (aqIdx - aqReIdxPred); bitCountFreq += entrCoder.indexGetBitCount (aqIdx - aqReIdxPred);
bitCountTime += entrCoder.indexGetBitCount (aqIdx - int (aqReIdxPrvGrp[b] & 31)); bitCountTime += entrCoder.indexGetBitCount (aqIdx - getPredCoefPrevGrp (aqReIdxPrvGrp[b]));
aqReIdxPred = aqIdx; aqReIdxPred = aqIdx;
@ -43,7 +48,7 @@ static uint32_t getDeltaCodeTimeFlag (const uint8_t* const alphaQCurr, const uns
aqIdx = gCplxPredUsed[b + 1] & 31; // TODO: <32 kHz short aqIdx = gCplxPredUsed[b + 1] & 31; // TODO: <32 kHz short
bitCountFreq += entrCoder.indexGetBitCount (aqIdx - aqImIdxPred); bitCountFreq += entrCoder.indexGetBitCount (aqIdx - aqImIdxPred);
bitCountTime += entrCoder.indexGetBitCount (aqIdx - int (aqImIdxPrvGrp[b] & 31)); bitCountTime += entrCoder.indexGetBitCount (aqIdx - getPredCoefPrevGrp (aqImIdxPrvGrp[b]));
aqImIdxPred = aqIdx; aqImIdxPred = aqIdx;
} }
@ -403,7 +408,7 @@ unsigned BitStreamWriter::writeStereoCoreToolInfo (const CoreCoderData& elData,
if (gCplxPredUsed[b] > 0) // write dpcm_alpha_q_re/_q_im if (gCplxPredUsed[b] > 0) // write dpcm_alpha_q_re/_q_im
{ {
int aqIdx = gCplxPredUsed[b] & 31; // range -15,...,15 int aqIdx = gCplxPredUsed[b] & 31; // range -15,...,15
int aqIdxDpcm = aqIdx - (deltaCodeTime > 0 ? int (aqReIdxPrvGrp[b] & 31) : aqReIdxPred); int aqIdxDpcm = aqIdx - (deltaCodeTime > 0 ? getPredCoefPrevGrp (aqReIdxPrvGrp[b]) : aqReIdxPred);
unsigned bits = entrCoder.indexGetBitCount (aqIdxDpcm); unsigned bits = entrCoder.indexGetBitCount (aqIdxDpcm);
if (deltaCodeTime == 0) aqReIdxPred = aqIdx; if (deltaCodeTime == 0) aqReIdxPred = aqIdx;
@ -413,7 +418,7 @@ unsigned BitStreamWriter::writeStereoCoreToolInfo (const CoreCoderData& elData,
if (complexCoef) if (complexCoef)
{ {
aqIdx = gCplxPredUsed[b + 1] & 31; // <32 kHz short! aqIdx = gCplxPredUsed[b + 1] & 31; // <32 kHz short!
aqIdxDpcm = aqIdx - (deltaCodeTime > 0 ? int (aqImIdxPrvGrp[b] & 31) : aqImIdxPred); aqIdxDpcm = aqIdx - (deltaCodeTime > 0 ? getPredCoefPrevGrp (aqImIdxPrvGrp[b]) : aqImIdxPred);
bits = entrCoder.indexGetBitCount (aqIdxDpcm); bits = entrCoder.indexGetBitCount (aqIdxDpcm);
if (deltaCodeTime == 0) aqImIdxPred = aqIdx; if (deltaCodeTime == 0) aqImIdxPred = aqIdx;

81
src/lib/exhaleEnc.cpp
View File

@ -252,6 +252,34 @@ static inline uint32_t getComplexRmsValue (const uint32_t rmsValue, const unsign
} }
#endif #endif
static inline uint8_t stereoCorrGrouping (const SfbGroupData& grpData, const unsigned nSamplesInFrame, uint8_t* stereoCorrData)
{
const uint16_t numWinGroup = grpData.numWindowGroups;
const uint16_t numBandsWin = nSamplesInFrame >> (SA_BW_SHIFT + 3);
uint32_t m = 0, w;
for (uint16_t gr = 0; gr < numWinGroup; gr++)
{
const uint16_t grpLength = grpData.windowGroupLength[gr];
const uint16_t grpLenFac = oneTwentyEightOver[grpLength]; // for grpStereoCorr/grpLength
for (uint16_t b = 0; b < numBandsWin; b++)
{
uint32_t grpStereoCorr = 0;
for (w = 0; w < grpLength; w++) grpStereoCorr += stereoCorrData[b + w * numBandsWin];
if (b == 0) m += grpStereoCorr;
grpStereoCorr = (grpStereoCorr * grpLenFac) >> 7;
for (w = 0; w < grpLength; w++) stereoCorrData[b + w * numBandsWin] = grpStereoCorr;
}
stereoCorrData += grpLength * numBandsWin;
}
return uint8_t ((m + 4) >> 3); // mean low-band correlation value
}
// ISO/IEC 23003-3, Table 75 // ISO/IEC 23003-3, Table 75
static inline unsigned toFrameLength (const USAC_CCFL coreCoderFrameLength) static inline unsigned toFrameLength (const USAC_CCFL coreCoderFrameLength)
{ {
@ -709,16 +737,18 @@ unsigned ExhaleEncoder::psychBitAllocation () // perceptual bit-allocation via s
} }
else // SCE or CPE: bandwidth-to-max_sfb mapping, short-window grouping for each channel else // SCE or CPE: bandwidth-to-max_sfb mapping, short-window grouping for each channel
{ {
// if ((coreConfig.stereoMode == 0) && (m_perCorrCurr[el] > SCHAR_MAX)) coreConfig.stereoMode = 1; if (coreConfig.commonWindow && (coreConfig.stereoMode == 0) && (m_perCorrHCurr[el] > SCHAR_MAX || m_perCorrLCurr[el] > (UCHAR_MAX * 3) / 4))
{
if (coreConfig.commonWindow && (m_perCorrCurr[el] > 128)) // run stereo pre-processing coreConfig.stereoMode = 1;
}
if (m_perCorrHCurr[el] > 128) // execute stereo pre-processing to increase correlation
{ {
const int16_t chanCorrSign = (coreConfig.stereoConfig & 2 ? -1 : 1); const int16_t chanCorrSign = (coreConfig.stereoConfig & 2 ? -1 : 1);
const bool eightShorts = (coreConfig.icsInfoCurr[0].windowSequence == EIGHT_SHORT); const bool eightShorts = (coreConfig.icsInfoCurr[0].windowSequence == EIGHT_SHORT);
const uint16_t nSamplesMax = (samplingRate < 37566 ? nSamplesInFrame : swbOffsetsL[m_swbTableIdx][maxSfbLong]); const uint16_t nSamplesMax = (samplingRate < 37566 ? nSamplesInFrame : swbOffsetsL[m_swbTableIdx][maxSfbLong]);
const uint8_t steppFadeLen = (eightShorts ? 4 : (coreConfig.tnsActive ? 32 : 64)); const uint8_t steppFadeLen = (eightShorts ? 4 : (coreConfig.tnsActive ? 32 : 64));
const uint8_t steppFadeOff = ((m_bitRateMode + 77000 / samplingRate) & 6) << (eightShorts ? 2 : 5); const uint8_t steppFadeOff = ((m_bitRateMode + 77000 / samplingRate) & 6) << (eightShorts ? 2 : 5);
const int64_t steppWeightI = __min (64, m_perCorrCurr[el] - 128) >> (eightShorts || coreConfig.tnsActive ? 1 : 0); const int64_t steppWeightI = __min (64, m_perCorrHCurr[el] - 128) >> (eightShorts || coreConfig.tnsActive ? 1 : 0);
const int64_t steppWeightD = 128 - steppWeightI; // decrement, (1 - crosstalk) * 128 const int64_t steppWeightD = 128 - steppWeightI; // decrement, (1 - crosstalk) * 128
const TnsData& tnsData0 = coreConfig.tnsData[0]; const TnsData& tnsData0 = coreConfig.tnsData[0];
const TnsData& tnsData1 = coreConfig.tnsData[1]; const TnsData& tnsData1 = coreConfig.tnsData[1];
@ -783,20 +813,21 @@ unsigned ExhaleEncoder::psychBitAllocation () // perceptual bit-allocation via s
} }
} // if coreConfig.commonWindow } // if coreConfig.commonWindow
if ((errorValue == 0) && (coreConfig.stereoMode == 2)) // frame M/S, synch statistics if ((errorValue == 0) && (coreConfig.stereoMode > 0)) // perform M/S, synch statistics
{ {
const uint8_t numSwbFrame = (coreConfig.icsInfoCurr[0].windowSequence == EIGHT_SHORT ? m_numSwbShort : __min (m_numSwbLong, maxSfbLong + 1)); const uint8_t numSwbFrame = (coreConfig.icsInfoCurr[0].windowSequence == EIGHT_SHORT ? m_numSwbShort : __min (m_numSwbLong, maxSfbLong + 1));
const uint32_t peakIndexSte = __max ((m_specAnaCurr[ci] >> 5) & 2047, (m_specAnaCurr[ci + 1] >> 5) & 2047) << 5; const uint32_t peakIndexSte = __max ((m_specAnaCurr[ci] >> 5) & 2047, (m_specAnaCurr[ci + 1] >> 5) & 2047) << 5;
errorValue = m_stereoCoder.applyFullFrameMatrix (m_mdctSignals[ci], m_mdctSignals[ci + 1], errorValue = m_stereoCoder.applyPredJointStereo (m_mdctSignals[ci], m_mdctSignals[ci + 1],
m_mdstSignals[ci], m_mdstSignals[ci + 1], m_mdstSignals[ci], m_mdstSignals[ci + 1],
coreConfig.groupingData[0], coreConfig.groupingData[1], coreConfig.groupingData[0], coreConfig.groupingData[1],
coreConfig.tnsData[0], coreConfig.tnsData[1], coreConfig.tnsData[0], coreConfig.tnsData[1],
numSwbFrame, coreConfig.stereoDataCurr, numSwbFrame, coreConfig.stereoDataCurr,
coreConfig.stereoConfig >> 1, coreConfig.stereoConfig & 1, m_bitRateMode <= 4, coreConfig.stereoMode > 1,
(coreConfig.stereoConfig >> 1) > 0, (coreConfig.stereoConfig & 1) > 0,
&sfbStepSizes[m_numSwbShort * NUM_WINDOW_GROUPS * ci], &sfbStepSizes[m_numSwbShort * NUM_WINDOW_GROUPS * ci],
&sfbStepSizes[m_numSwbShort * NUM_WINDOW_GROUPS * (ci + 1)]); &sfbStepSizes[m_numSwbShort * NUM_WINDOW_GROUPS * (ci + 1)]);
if (errorValue == 2) // use frame M/S with cplx_pred_all=1 if (errorValue == 2) // signal M/S with complex prediction
{ {
coreConfig.stereoMode += 2; errorValue = 0; coreConfig.stereoMode += 2; errorValue = 0;
} }
@ -805,6 +836,7 @@ unsigned ExhaleEncoder::psychBitAllocation () // perceptual bit-allocation via s
meanSpecFlat[ci] = meanSpecFlat[ci + 1] = ((uint16_t) meanSpecFlat[ci] + (uint16_t) meanSpecFlat[ci + 1]) >> 1; meanSpecFlat[ci] = meanSpecFlat[ci + 1] = ((uint16_t) meanSpecFlat[ci] + (uint16_t) meanSpecFlat[ci + 1]) >> 1;
// meanTempFlat[ci] = meanTempFlat[ci + 1] = ((uint16_t) meanTempFlat[ci] + (uint16_t) meanTempFlat[ci + 1]) >> 1; // meanTempFlat[ci] = meanTempFlat[ci + 1] = ((uint16_t) meanTempFlat[ci] + (uint16_t) meanTempFlat[ci + 1]) >> 1;
} }
else memset (coreConfig.stereoDataCurr, 0, (MAX_NUM_SWB_SHORT * NUM_WINDOW_GROUPS) * sizeof (uint8_t));
for (unsigned ch = 0; ch < nrChannels; ch++) // channel loop for (unsigned ch = 0; ch < nrChannels; ch++) // channel loop
{ {
@ -824,6 +856,7 @@ unsigned ExhaleEncoder::psychBitAllocation () // perceptual bit-allocation via s
const uint32_t* refRms = &coreConfig.groupingData[1 - ch].sfbRmsValues[m_numSwbShort * gr]; const uint32_t* refRms = &coreConfig.groupingData[1 - ch].sfbRmsValues[m_numSwbShort * gr];
uint8_t* grpScaleFacs = &grpData.scaleFactors[m_numSwbShort * gr]; uint8_t* grpScaleFacs = &grpData.scaleFactors[m_numSwbShort * gr];
uint32_t* grpStepSizes = &stepSizes[m_numSwbShort * gr]; uint32_t* grpStepSizes = &stepSizes[m_numSwbShort * gr];
uint8_t* grpStereoData = &coreConfig.stereoDataCurr[m_numSwbShort * gr];
uint32_t b, grpRmsMin = INT_MAX; // min. RMS value, used for overcoding reduction uint32_t b, grpRmsMin = INT_MAX; // min. RMS value, used for overcoding reduction
// undercoding reduction for case where large number of coefs is quantized to zero // undercoding reduction for case where large number of coefs is quantized to zero
@ -831,7 +864,7 @@ unsigned ExhaleEncoder::psychBitAllocation () // perceptual bit-allocation via s
for (b = 0; b < maxSfbCh; b++) for (b = 0; b < maxSfbCh; b++)
{ {
#if SA_IMPROVED_REAL_ABS #if SA_IMPROVED_REAL_ABS
const uint32_t rmsComp = (coreConfig.stereoMode > 0 ? squareMeanRoot (refRms[b], grpRms[b]) : grpRms[b]); const uint32_t rmsComp = (grpStereoData[b] > 0 ? squareMeanRoot (refRms[b], grpRms[b]) : grpRms[b]);
const uint32_t rmsRef9 = (coreConfig.commonWindow ? refRms[b] >> 9 : rmsComp); const uint32_t rmsRef9 = (coreConfig.commonWindow ? refRms[b] >> 9 : rmsComp);
#else #else
const uint32_t rmsComp = getComplexRmsValue (grpRms[b], gr, b, numSwbCh, coreConfig.tnsData[ch]); const uint32_t rmsComp = getComplexRmsValue (grpRms[b], gr, b, numSwbCh, coreConfig.tnsData[ch]);
@ -847,7 +880,7 @@ unsigned ExhaleEncoder::psychBitAllocation () // perceptual bit-allocation via s
if ((samplingRate >= 27713) && (b < maxSfbLong) && !eightShorts) // uncoded coefs if ((samplingRate >= 27713) && (b < maxSfbLong) && !eightShorts) // uncoded coefs
{ {
#if SA_IMPROVED_REAL_ABS #if SA_IMPROVED_REAL_ABS
const uint32_t rmsComp = (coreConfig.stereoMode > 0 ? squareMeanRoot (refRms[b], grpRms[b]) : grpRms[b]); const uint32_t rmsComp = (grpStereoData[b] > 0 ? squareMeanRoot (refRms[b], grpRms[b]) : grpRms[b]);
const uint32_t rmsRef9 = (coreConfig.commonWindow ? refRms[b] >> 9 : rmsComp); const uint32_t rmsRef9 = (coreConfig.commonWindow ? refRms[b] >> 9 : rmsComp);
#else #else
const uint32_t rmsComp = getComplexRmsValue (grpRms[b], gr, b, numSwbCh, coreConfig.tnsData[ch]); const uint32_t rmsComp = getComplexRmsValue (grpRms[b], gr, b, numSwbCh, coreConfig.tnsData[ch]);
@ -867,7 +900,7 @@ unsigned ExhaleEncoder::psychBitAllocation () // perceptual bit-allocation via s
const unsigned lfAtten = (b <= 5 ? (eightShorts ? 1 : 4) + b * lfConst : 5 * lfConst - 1 + b + ((b + 5) >> 4)); const unsigned lfAtten = (b <= 5 ? (eightShorts ? 1 : 4) + b * lfConst : 5 * lfConst - 1 + b + ((b + 5) >> 4));
const uint8_t sfbWidth = grpOff[b + 1] - grpOff[b]; const uint8_t sfbWidth = grpOff[b + 1] - grpOff[b];
const uint64_t rateFac = mSfmFac * s * __min (32, lfAtten * grpData.numWindowGroups); // rate control part 1 const uint64_t rateFac = mSfmFac * s * __min (32, lfAtten * grpData.numWindowGroups); // rate control part 1
const uint64_t sScaled = ((1u << 23) + __max (grpRmsMin, grpStepSizes[b]) * (scaleBr - (coreConfig.stereoMode > 0 ? 1 : 0)) * rateFac) >> 24; const uint64_t sScaled = ((1u << 23) + __max (grpRmsMin, grpStepSizes[b]) * (scaleBr - (grpStereoData[b] > 0 ? 1 : 0)) * rateFac) >> 24;
// scale step-sizes according to VBR mode & derive scale factors from step-sizes // scale step-sizes according to VBR mode & derive scale factors from step-sizes
grpStepSizes[b] = uint32_t (__max (BA_EPS, __min (UINT_MAX, sScaled))); grpStepSizes[b] = uint32_t (__max (BA_EPS, __min (UINT_MAX, sScaled)));
@ -1224,23 +1257,24 @@ unsigned ExhaleEncoder::spectralProcessing () // complete ics_info(), calc TNS
nSamplesInFrame, eightShorts, coreConfig.stereoDataCurr); nSamplesInFrame, eightShorts, coreConfig.stereoDataCurr);
if (steAnaStats == SHRT_MIN) errorValue = 1; if (steAnaStats == SHRT_MIN) errorValue = 1;
if ((s = abs (steAnaStats)) * m_perCorrCurr[el] == 0) // transitions to/from silence if ((s = abs (steAnaStats)) * m_perCorrHCurr[el] == 0) // transition to/from silence
{ {
m_perCorrCurr[el] = uint8_t((32 + s * __min (64, eightTimesSqrt256Minus[meanSpecFlat])) >> 6); m_perCorrHCurr[el] = uint8_t((32 + s * __min (64, eightTimesSqrt256Minus[meanSpecFlat])) >> 6);
} }
else // gentle overlap length dependent temporal smoothing else // gentle overlap length dependent temporal smoothing
{ {
const int16_t allowedDiff = (coreConfig.icsInfoCurr[0].windowSequence < EIGHT_SHORT ? 16 : 32); const int16_t allowedDiff = (coreConfig.icsInfoCurr[0].windowSequence < EIGHT_SHORT ? 16 : 32);
const int16_t prevPerCorr = __max (128, __min (192, m_perCorrCurr[el])); const int16_t prevPerCorr = __max (128, __min (192, m_perCorrHCurr[el]));
const int16_t currPerCorr = (32 + s * __min (64, eightTimesSqrt256Minus[meanSpecFlat])) >> 6; const int16_t currPerCorr = (32 + s * __min (64, eightTimesSqrt256Minus[meanSpecFlat])) >> 6;
m_perCorrCurr[el] = (uint8_t) __max (prevPerCorr - allowedDiff, __min (prevPerCorr + allowedDiff, currPerCorr)); m_perCorrHCurr[el] = (uint8_t) __max (prevPerCorr - allowedDiff, __min (prevPerCorr + allowedDiff, currPerCorr));
} }
m_perCorrLCurr[el] = coreConfig.stereoDataCurr[0];
if ((int) s == steAnaStats * -1) coreConfig.stereoConfig = 2; // 2: S>M, pred_dir=1 if ((int) s == steAnaStats * -1) coreConfig.stereoConfig = 2; // 2: S>M, pred_dir=1
if (s > (UCHAR_MAX * 3) / 4) coreConfig.stereoMode = 2; // 2: all, ms_mask_present=2 if (s > (UCHAR_MAX * 3) / 4) coreConfig.stereoMode = 2; // 2: all, ms_mask_present=2
} }
else if (nrChannels > 1) m_perCorrCurr[el] = 128; // update history with halfway value else if (nrChannels > 1) m_perCorrHCurr[el] = m_perCorrLCurr[el] = 128; // "mid" value
for (unsigned ch = 0; ch < nrChannels; ch++) // channel loop for (unsigned ch = 0; ch < nrChannels; ch++) // channel loop
{ {
@ -1323,7 +1357,7 @@ unsigned ExhaleEncoder::spectralProcessing () // complete ics_info(), calc TNS
uint8_t& maxSfb1 = coreConfig.icsInfoCurr[1].maxSfb; uint8_t& maxSfb1 = coreConfig.icsInfoCurr[1].maxSfb;
const uint8_t maxSfbSte = __max (maxSfb0, maxSfb1); // max_sfb_ste, as in Table 24 const uint8_t maxSfbSte = __max (maxSfb0, maxSfb1); // max_sfb_ste, as in Table 24
if ((maxSfb0 > 0) && (maxSfb1 > 0) && (maxSfbSte - __min (maxSfb0, maxSfb1) <= 1 || coreConfig.stereoMode == 2)) if ((maxSfb0 > 0) && (maxSfb1 > 0) && (maxSfbSte - __min (maxSfb0, maxSfb1) <= 1 || coreConfig.stereoMode > 0))
{ {
uint32_t& sa0 = m_specAnaCurr[ci-2]; uint32_t& sa0 = m_specAnaCurr[ci-2];
uint32_t& sa1 = m_specAnaCurr[ci-1]; uint32_t& sa1 = m_specAnaCurr[ci-1];
@ -1367,6 +1401,11 @@ unsigned ExhaleEncoder::spectralProcessing () // complete ics_info(), calc TNS
} }
} }
maxSfb0 = maxSfb1 = maxSfbSte; maxSfb0 = maxSfb1 = maxSfbSte;
if ((m_bitRateMode <= 4) && (coreConfig.icsInfoCurr[0].windowSequence == EIGHT_SHORT))
{
m_perCorrLCurr[el] = stereoCorrGrouping (coreConfig.groupingData[0], nSamplesInFrame, coreConfig.stereoDataCurr);
}
} }
else coreConfig.stereoMode = 0; // since a max_sfb is 0 else coreConfig.stereoMode = 0; // since a max_sfb is 0
@ -1583,14 +1622,15 @@ unsigned ExhaleEncoder::temporalProcessing () // determine time-domain aspects o
} }
coreConfig.commonWindow = (winSeq0 == winSeq1); // synch coreConfig.commonWindow = (winSeq0 == winSeq1); // synch
memset (coreConfig.stereoDataPrev, 16, (MAX_NUM_SWB_LONG + 1) * sizeof (uint8_t)); memset (coreConfig.stereoDataPrev, 0, (MAX_NUM_SWB_LONG + 1) * sizeof (uint8_t));
if (((winSeq0 == EIGHT_SHORT) == (coreConfig.icsInfoPrev[0].windowSequence == EIGHT_SHORT)) && !m_indepFlag && if (((winSeq0 == EIGHT_SHORT) == (coreConfig.icsInfoPrev[0].windowSequence == EIGHT_SHORT)) && !m_indepFlag &&
((winSeq1 == EIGHT_SHORT) == (coreConfig.icsInfoPrev[1].windowSequence == EIGHT_SHORT)) && (coreConfig.stereoMode > 0)) ((winSeq1 == EIGHT_SHORT) == (coreConfig.icsInfoPrev[1].windowSequence == EIGHT_SHORT)) && (coreConfig.stereoMode > 0))
{ {
const unsigned lastGrpOffset = (coreConfig.icsInfoPrev[0].windowSequence == EIGHT_SHORT ? m_numSwbShort * (NUM_WINDOW_GROUPS - 1) : 0); const unsigned lastGrpOffset = (coreConfig.icsInfoPrev[0].windowSequence == EIGHT_SHORT ? m_numSwbShort * (NUM_WINDOW_GROUPS - 1) : 0);
const unsigned maxSfbStePrev = __max (coreConfig.icsInfoPrev[0].maxSfb, coreConfig.icsInfoPrev[1].maxSfb);
memcpy (coreConfig.stereoDataPrev, &coreConfig.stereoDataCurr[lastGrpOffset], (coreConfig.icsInfoPrev[0].maxSfb + 1) * sizeof (uint8_t)); memcpy (coreConfig.stereoDataPrev, &coreConfig.stereoDataCurr[lastGrpOffset], (maxSfbStePrev + 1) * sizeof (uint8_t));
} }
} // if nrChannels > 1 } // if nrChannels > 1
} }
@ -1656,7 +1696,8 @@ ExhaleEncoder::ExhaleEncoder (int32_t* const inputPcmData, unsigned ch
const ELEM_TYPE et = elementTypeConfig[m_channelConf % USAC_MAX_NUM_ELCONFIGS][el]; // usacElementType const ELEM_TYPE et = elementTypeConfig[m_channelConf % USAC_MAX_NUM_ELCONFIGS][el]; // usacElementType
m_elementData[el] = nullptr; m_elementData[el] = nullptr;
m_perCorrCurr[el] = 0; m_perCorrHCurr[el] = 0;
m_perCorrLCurr[el] = 0;
#if !RESTRICT_TO_AAC #if !RESTRICT_TO_AAC
m_noiseFilling[el] = (useNoiseFilling && (et < ID_USAC_LFE)); m_noiseFilling[el] = (useNoiseFilling && (et < ID_USAC_LFE));
m_timeWarping[el] = (false /* N/A */ && (et < ID_USAC_LFE)); m_timeWarping[el] = (false /* N/A */ && (et < ID_USAC_LFE));

3
src/lib/exhaleEnc.h
View File

@ -89,7 +89,8 @@ private:
unsigned char* m_outAuData; unsigned char* m_outAuData;
BitStreamWriter m_outStream; // for access unit creation BitStreamWriter m_outStream; // for access unit creation
int32_t* m_pcm24Data; int32_t* m_pcm24Data;
uint8_t m_perCorrCurr[USAC_MAX_NUM_ELEMENTS]; uint8_t m_perCorrHCurr[USAC_MAX_NUM_ELEMENTS];
uint8_t m_perCorrLCurr[USAC_MAX_NUM_ELEMENTS];
SfbGroupData* m_scaleFacData[USAC_MAX_NUM_CHANNELS]; SfbGroupData* m_scaleFacData[USAC_MAX_NUM_CHANNELS];
SfbQuantizer m_sfbQuantizer; // powerlaw quantization SfbQuantizer m_sfbQuantizer; // powerlaw quantization
SpecAnalyzer m_specAnalyzer; // for spectral analysis SpecAnalyzer m_specAnalyzer; // for spectral analysis

20
src/lib/stereoProcessing.cpp
View File

@ -18,19 +18,21 @@ StereoProcessor::StereoProcessor ()
} }
// public functions // public functions
unsigned StereoProcessor::applyFullFrameMatrix (int32_t* const mdctSpectrum1, int32_t* const mdctSpectrum2, unsigned StereoProcessor::applyPredJointStereo (int32_t* const mdctSpectrum1, int32_t* const mdctSpectrum2,
int32_t* const mdstSpectrum1, int32_t* const mdstSpectrum2, int32_t* const mdstSpectrum1, int32_t* const mdstSpectrum2,
SfbGroupData& groupingData1, SfbGroupData& groupingData2, SfbGroupData& groupingData1, SfbGroupData& groupingData2,
const TnsData& filterData1, const TnsData& filterData2, const TnsData& filterData1, const TnsData& filterData2,
const uint8_t numSwbFrame, uint8_t* const sfbStereoData, const uint8_t numSwbFrame, uint8_t* const sfbStereoData,
const uint8_t useAltPredDir, const uint8_t useComplexCoef, const bool usePerCorrData, const bool useFullFrameMS,
const bool reversePredDir, const bool useComplexCoef,
uint32_t* const sfbStepSize1, uint32_t* const sfbStepSize2) uint32_t* const sfbStepSize1, uint32_t* const sfbStepSize2)
{ {
const bool applyPredSte = (sfbStereoData != nullptr); // use real-valued predictive stereo const bool applyPredSte = (sfbStereoData != nullptr); // use real-valued predictive stereo
const bool alterPredDir = (applyPredSte && (useAltPredDir > 0)); // predict mid from side? const bool alterPredDir = (applyPredSte && reversePredDir); // predict mid from side band?
const SfbGroupData& grp = groupingData1; const SfbGroupData& grp = groupingData1;
const bool eightShorts = (grp.numWindowGroups > 1); const bool eightShorts = (grp.numWindowGroups > 1);
const uint8_t maxSfbSte = (eightShorts ? __min (numSwbFrame, __max (grp.sfbsPerGroup, groupingData2.sfbsPerGroup) + 1) : numSwbFrame); const uint8_t maxSfbSte = (eightShorts ? __min (numSwbFrame, __max (grp.sfbsPerGroup, groupingData2.sfbsPerGroup) + 1) : numSwbFrame);
const bool perCorrData = (usePerCorrData && !eightShorts); // use perceptual correlation?
uint32_t numSfbPredSte = 0; // counter uint32_t numSfbPredSte = 0; // counter
if ((mdctSpectrum1 == nullptr) || (mdctSpectrum2 == nullptr) || (numSwbFrame < maxSfbSte) || (grp.numWindowGroups != groupingData2.numWindowGroups) || if ((mdctSpectrum1 == nullptr) || (mdctSpectrum2 == nullptr) || (numSwbFrame < maxSfbSte) || (grp.numWindowGroups != groupingData2.numWindowGroups) ||
@ -38,8 +40,16 @@ unsigned StereoProcessor::applyFullFrameMatrix (int32_t* const mdctSpectrum1, in
{ {
return 1; // invalid arguments error return 1; // invalid arguments error
} }
#if 1
if (!useFullFrameMS) // TODO
{
if (applyPredSte) memset (sfbStereoData, 0, (MAX_NUM_SWB_SHORT * NUM_WINDOW_GROUPS) * sizeof (uint8_t));
if (applyPredSte && !eightShorts) memcpy (m_stereoCorrValue, sfbStereoData, (grp.sfbOffsets[numSwbFrame] >> SA_BW_SHIFT) * sizeof (uint8_t)); return 0;
}
#endif
if (applyPredSte && perCorrData) memcpy (m_stereoCorrValue, sfbStereoData, (grp.sfbOffsets[numSwbFrame] >> SA_BW_SHIFT) * sizeof (uint8_t));
for (uint16_t gr = 0; gr < grp.numWindowGroups; gr++) for (uint16_t gr = 0; gr < grp.numWindowGroups; gr++)
{ {
@ -277,7 +287,7 @@ unsigned StereoProcessor::applyFullFrameMatrix (int32_t* const mdctSpectrum1, in
sfbStereoData[sfbEv + 1 + numSwbFrame * gr] = uint8_t (b + 16); // init alpha_q_im sfbStereoData[sfbEv + 1 + numSwbFrame * gr] = uint8_t (b + 16); // init alpha_q_im
#endif #endif
if (!eightShorts && ((offEv & (SA_BW - 1)) == 0) && ((width & (SA_BW - 1)) == 0)) if (perCorrData && ((offEv & (SA_BW - 1)) == 0) && ((width & (SA_BW - 1)) == 0))
{ {
const uint8_t* const perCorr = &m_stereoCorrValue[offEv >> SA_BW_SHIFT]; const uint8_t* const perCorr = &m_stereoCorrValue[offEv >> SA_BW_SHIFT];

5
src/lib/stereoProcessing.h
View File

@ -34,12 +34,13 @@ public:
// destructor // destructor
~StereoProcessor () { } ~StereoProcessor () { }
// public functions // public functions
unsigned applyFullFrameMatrix (int32_t* const mdctSpectrum1, int32_t* const mdctSpectrum2, unsigned applyPredJointStereo (int32_t* const mdctSpectrum1, int32_t* const mdctSpectrum2,
int32_t* const mdstSpectrum1, int32_t* const mdstSpectrum2, int32_t* const mdstSpectrum1, int32_t* const mdstSpectrum2,
SfbGroupData& groupingData1, SfbGroupData& groupingData2, SfbGroupData& groupingData1, SfbGroupData& groupingData2,
const TnsData& filterData1, const TnsData& filterData2, const TnsData& filterData1, const TnsData& filterData2,
const uint8_t numSwbFrame, uint8_t* const sfbStereoData, const uint8_t numSwbFrame, uint8_t* const sfbStereoData,
const uint8_t useAltPredDir, const uint8_t useComplexCoef, const bool usePerCorrData, const bool useFullFrameMS,
const bool reversePredDir, const bool useComplexCoef,
uint32_t* const sfbStepSize1, uint32_t* const sfbStepSize2); uint32_t* const sfbStepSize1, uint32_t* const sfbStepSize2);
}; // StereoProcessor }; // StereoProcessor