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
Files
057bb87e64901bd19502a3d13e4556c80c64c2f0
exhale/src/lib/bitAllocation.cpp
Christian R. Helmrich 057bb87e64 clean DTX and RC
2021-06-28 23:00:00 +02:00

536 lines
24 KiB
C++
Raw Blame History

/* bitAllocation.cpp - source file for class needed for psychoacoustic bit-allocation
* written by C. R. Helmrich, last modified in 2021 - see License.htm for legal notices
*
* The copyright in this software is being made available under the exhale Copyright License
* and comes with ABSOLUTELY NO WARRANTY. This software may be subject to other third-
* party rights, including patent rights. No such rights are granted under this License.
*
* Copyright (c) 2018-2021 Christian R. Helmrich, project ecodis. All rights reserved.
*/
#include "exhaleLibPch.h"
#include "bitAllocation.h"
// static helper functions
static inline uint32_t intSqrt (const uint32_t val)
{
return uint32_t (0.5 + sqrt ((double) val));
}
static inline uint32_t jndModel (const uint32_t val, const uint32_t mean,
const unsigned expTimes512, const unsigned mulTimes512)
{
const double exp = (double) expTimes512 / 512.0; // exponent
const double mul = (double) mulTimes512 / 512.0; // a factor
const double res = pow (mul * (double) val, exp) * pow ((double) mean, 1.0 - exp);
return uint32_t (__min ((double) UINT_MAX, res + 0.5));
}
static inline uint32_t squareMeanRoot (const uint32_t value1, const uint32_t value2)
{
const double meanRoot = (sqrt ((double) value1) + sqrt ((double) value2)) * 0.5;
return uint32_t (meanRoot * meanRoot + 0.5);
}
static void jndPowerLawAndPeakSmoothing (uint32_t* const stepSizes, const unsigned nStepSizes, const bool lowRateMode,
const uint32_t avgStepSize, const uint8_t sfm, const uint8_t tfm)
{
const unsigned expTimes512 = 512u - sfm; // 1.0 - sfm / 2.0
const unsigned mulTimes512 = __min (expTimes512, 512u - tfm);
uint32_t stepSizeM3 = 0, stepSizeM2 = 0, stepSizeM1 = BA_EPS; // hearing threshold around zero Hz
unsigned b;
for (b = 0; b < __min (2, nStepSizes); b++)
{
stepSizeM3 = stepSizeM2;
stepSizeM2 = stepSizeM1;
stepSizeM1 = stepSizes[b] = jndModel (stepSizes[b], avgStepSize, expTimes512, mulTimes512);
}
stepSizes[0] = __min (stepSizeM1, stepSizes[0]); // `- becomes --
for (/*b*/; b < nStepSizes; b++)
{
const uint64_t modifiedB = (lowRateMode ? 16 + b : b);
const uint64_t oneMinusB = 128 - modifiedB;
const uint32_t stepSizeB = jndModel (stepSizes[b], avgStepSize, expTimes512, mulTimes512);
if ((stepSizeM3 <= stepSizeM2) && (stepSizeM3 <= stepSizeM1) && (stepSizeB <= stepSizeM2) && (stepSizeB <= stepSizeM1))
{
const uint32_t maxM3M0 = __max (stepSizeM3, stepSizeB); // smoothen local spectral peak of _<>`- shape
stepSizes[b - 2] = uint32_t ((modifiedB * stepSizes[b - 2] + oneMinusB * __min (maxM3M0, stepSizes[b - 2]) + 64) >> 7); // _-`-
stepSizes[b - 1] = uint32_t ((modifiedB * stepSizes[b - 1] + oneMinusB * __min (maxM3M0, stepSizes[b - 1]) + 64) >> 7); // _---
}
stepSizeM3 = stepSizeM2;
stepSizeM2 = stepSizeM1;
stepSizeM1 = (stepSizes[b] = stepSizeB); // modified step-size may be smoothened in next loop iteration
}
}
// constructor
BitAllocator::BitAllocator ()
{
for (unsigned ch = 0; ch < USAC_MAX_NUM_CHANNELS; ch++)
{
m_avgStepSize[ch] = 0;
m_avgSpecFlat[ch] = 0;
m_avgTempFlat[ch] = 0;
}
m_rateIndex = 0;
m_tempSfbValue = nullptr;
m_tnsPredictor = nullptr;
}
// public functions
void BitAllocator::getChAverageSpecFlat (uint8_t meanSpecFlatInCh[USAC_MAX_NUM_CHANNELS], const unsigned nChannels)
{
if ((meanSpecFlatInCh == nullptr) || (nChannels > USAC_MAX_NUM_CHANNELS))
{
return;
}
memcpy (meanSpecFlatInCh, m_avgSpecFlat, nChannels * sizeof (uint8_t));
}
void BitAllocator::getChAverageTempFlat (uint8_t meanTempFlatInCh[USAC_MAX_NUM_CHANNELS], const unsigned nChannels)
{
if ((meanTempFlatInCh == nullptr) || (nChannels > USAC_MAX_NUM_CHANNELS))
{
return;
}
memcpy (meanTempFlatInCh, m_avgTempFlat, nChannels * sizeof (uint8_t));
}
uint16_t BitAllocator::getRateCtrlFac (const int32_t rateRatio, const unsigned samplingRate, const uint32_t specFlatness)
{
const uint32_t brRatio = __max (1 << 15, __min (USHRT_MAX, rateRatio * (36 - 9 * m_rateIndex)));
const uint16_t mSfmSqr = (m_rateIndex < 2 && samplingRate >= 27713 ? (specFlatness * specFlatness) >> m_rateIndex : 0);
const uint16_t mSfmFac = 256 - (((32 + m_rateIndex) * (specFlatness << 4) - mSfmSqr + (1 << 9)) >> 10);
return uint16_t ((brRatio * mSfmFac + (1 << 7)) >> 8);
}
uint8_t BitAllocator::getScaleFac (const uint32_t sfbStepSize, const int32_t* const sfbSignal, const uint8_t sfbWidth,
const uint32_t sfbRmsValue)
{
uint8_t sf;
uint32_t u;
#if !RESTRICT_TO_AAC
uint64_t meanSpecLoudness = 0;
double d;
#endif
if ((sfbSignal == nullptr) || (sfbWidth == 0) || (sfbRmsValue < 46))
{
return 0; // use lowest scale factor
}
#if RESTRICT_TO_AAC
u = 0;
for (sf = 0; sf < sfbWidth; sf++)
{
u += intSqrt (abs (sfbSignal[sf]));
}
u = uint32_t ((u * 16384ui64 + (sfbWidth >> 1)) / sfbWidth);
u = uint32_t (0.5 + sqrt ((double) u) * 128.0);
if (u < 42567) return 0;
u = uint32_t ((sfbStepSize * 42567ui64 + (u >> 1)) / u);
sf = (u > 1 ? uint8_t (0.5 + 17.7169498394 * log10 ((double) u)) : 4);
#else
for (sf = 0; sf < sfbWidth; sf++) // simple, low-complexity derivation method for USAC's arithmetic coder
{
const int64_t temp = ((int64_t) sfbSignal[sf] + 8) >> 4; // avoid overflow
meanSpecLoudness += temp * temp;
}
meanSpecLoudness = uint64_t (0.5 + pow (256.0 * (double) meanSpecLoudness / sfbWidth, 0.25));
u = uint32_t (0.5 + pow ((double) sfbRmsValue, 0.75) * 256.0); // u = 2^8 * (sfbRmsValue^0.75)
u = uint32_t ((meanSpecLoudness * sfbStepSize * 665 + (u >> 1)) / u); // u = sqrt(6.75) * m*thr/u
d = (u > 1 ? log10 ((double) u) : 0.25);
u = uint32_t (0.5 + pow ((double) sfbRmsValue, 0.25) * 16384.0); // u = 2^14 * (sfbRmsValue^0.25)
u = uint32_t (((uint64_t) sfbStepSize * 42567 + (u >> 1)) / u); // u = sqrt(6.75) * thr/u
d += (u > 1 ? log10 ((double) u) : 0.25);
sf = uint8_t (0.5 + 8.8584749197 * d); // sf = (8/3) * log2(u1*u2) = (8/3) * (log2(u1)+log2(u2))
#endif
return __min (SCHAR_MAX, sf);
}
unsigned BitAllocator::initAllocMemory (LinearPredictor* const linPredictor, const uint8_t numSwb, const uint8_t bitRateMode)
{
if (linPredictor == nullptr)
{
return 1; // invalid arguments error
}
m_rateIndex = bitRateMode;
m_tnsPredictor = linPredictor;
if ((m_tempSfbValue = (uint8_t*) malloc (__max (MAX_PREDICTION_ORDER * sizeof (short), numSwb) * sizeof (uint8_t))) == nullptr)
{
return 2; // memory allocation error
}
return 0; // no error
}
unsigned BitAllocator::initSfbStepSizes (const SfbGroupData* const groupData[USAC_MAX_NUM_CHANNELS], const uint8_t numSwbShort,
const uint32_t specAnaStats[USAC_MAX_NUM_CHANNELS],
const uint32_t tempAnaStats[USAC_MAX_NUM_CHANNELS],
const unsigned nChannels, const unsigned samplingRate, uint32_t* const sfbStepSizes,
const unsigned lfeChannelIndex /*= USAC_MAX_NUM_CHANNELS*/, const bool tnsDisabled /*= false*/)
{
// equal-loudness weighting based on data from: K. Kurakata, T. Mizunami, and K. Matsushita, "Percentiles
// of Normal Hearing-Threshold Distribution Under Free-Field Listening Conditions in Numerical Form," Ac.
// Sci. Tech, vol. 26, no. 5, pp. 447-449, Jan. 2005, https://www.researchgate.net/publication/239433096.
const unsigned HF/*idx*/= ((123456 - samplingRate) >> 11) + (samplingRate < 37566 ? 2 : 0); // start SFB
const unsigned LF/*idx*/= 9;
const unsigned MF/*idx*/= (samplingRate < 27713 ? HF : __min (HF, 30u));
const unsigned msShift = (samplingRate + 36736) >> 15; // TODO: 768 smp
const unsigned msOffset = 1 << (msShift - 1);
uint32_t nMeans = 0, sumMeans = 0;
if ((groupData == nullptr) || (specAnaStats == nullptr) || (tempAnaStats == nullptr) || (sfbStepSizes == nullptr) ||
(numSwbShort < MIN_NUM_SWB_SHORT) || (numSwbShort > MAX_NUM_SWB_SHORT) || (nChannels > USAC_MAX_NUM_CHANNELS) ||
(samplingRate < 7350) || (samplingRate > 96000) || (lfeChannelIndex > USAC_MAX_NUM_CHANNELS))
{
return 1; // invalid arguments error
}
for (unsigned ch = 0; ch < nChannels; ch++)
{
const SfbGroupData& grpData = *groupData[ch];
const uint32_t maxSfbInCh = __min (MAX_NUM_SWB_LONG, grpData.sfbsPerGroup);
const uint32_t nBandsInCh = grpData.numWindowGroups * maxSfbInCh;
const uint32_t* rms = grpData.sfbRmsValues;
uint32_t* stepSizes = &sfbStepSizes[ch * numSwbShort * NUM_WINDOW_GROUPS];
// --- apply INTRA-channel simultaneous masking, equal-loudness weighting, and thresholding to SFB RMS data
uint32_t maskingSlope = 0, gr, b, elw = 58254; // 8/9
uint32_t rmsEqualLoud = 0;
uint32_t sumStepSizes = 0;
m_avgStepSize[ch] = 0;
b = ((specAnaStats[ch] >> 16) & UCHAR_MAX);
b = __max (b * b, (tempAnaStats[ch] >> 24) * (tempAnaStats[ch] >> 24));
m_avgSpecFlat[ch] = uint8_t ((b + (1 << 7)) >> 8); // max. of squared SFM from spec. and temp. analysis
b = ((tempAnaStats[ch] >> 16) & UCHAR_MAX);
b = __max (b * b, (specAnaStats[ch] >> 24) * (specAnaStats[ch] >> 24));
m_avgTempFlat[ch] = uint8_t ((b + (1 << 7)) >> 8); // max. of squared TFM from spec. and temp. analysis
if ((nBandsInCh == 0) || (grpData.numWindowGroups > NUM_WINDOW_GROUPS))
{
continue;
}
if ((ch == lfeChannelIndex) || (grpData.numWindowGroups != 1)) // LFE, SHORT windows: no masking or ELW
{
for (gr = 0; gr < grpData.numWindowGroups; gr++)
{
const uint32_t* gRms = &rms[numSwbShort * gr];
uint32_t* gStepSizes = &stepSizes[numSwbShort * gr];
for (b = numSwbShort - 1; b >= maxSfbInCh; b--)
{
gStepSizes[b] = 0;
}
for (/*b*/; b > 0; b--)
{
gStepSizes[b] = __max (gRms[b], BA_EPS);
sumStepSizes += intSqrt (gStepSizes[b]);
}
gStepSizes[0] = __max (gRms[0], BA_EPS);
sumStepSizes += intSqrt (gStepSizes[0]);
} // for gr
if (ch != lfeChannelIndex)
{
// --- SHORT windows: apply perceptual just noticeable difference (JND) model and local band-peak smoothing
nMeans++;
for (b = maxSfbInCh - 1; b > 0; b--) // gentle temporal band-peak smoothing; a spectral one follows
{
uint32_t maxGrpStep = stepSizes[b], stepSizeM1 = BA_EPS;
for (gr = 1; gr < grpData.numWindowGroups; gr++)
{
const uint32_t curGrpStep = stepSizes[b + numSwbShort * gr];
if (curGrpStep > maxGrpStep) maxGrpStep = curGrpStep;
}
for (gr = 0; gr < grpData.numWindowGroups; gr++)
{
const uint32_t newGrpStep = __max (stepSizeM1, (gr + 1 == grpData.numWindowGroups ? BA_EPS : stepSizes[b + numSwbShort * (gr + 1)]));
stepSizeM1 = stepSizes[b + numSwbShort * gr];
if ((stepSizeM1 == maxGrpStep) && (maxGrpStep > newGrpStep))
{
const uint32_t sqrtOldStep = intSqrt (maxGrpStep);
const uint32_t sqrtNewStep = intSqrt (newGrpStep);
uint32_t& gStepSize = stepSizes[b + numSwbShort * gr];
sumStepSizes += (gStepSize = (sqrtOldStep + sqrtNewStep) >> 1) - sqrtOldStep;
gStepSize *= gStepSize; // for square-mean-root
}
}
} // for b
m_avgStepSize[ch] = __min (USHRT_MAX, (sumStepSizes + (nBandsInCh >> 1)) / nBandsInCh);
sumMeans += m_avgStepSize[ch];
m_avgStepSize[ch] *= m_avgStepSize[ch];
for (gr = 0; gr < grpData.numWindowGroups; gr++) // separate spectral peak smoothing for each group
{
jndPowerLawAndPeakSmoothing (&stepSizes[numSwbShort * gr], maxSfbInCh, false,
m_avgStepSize[ch], m_avgSpecFlat[ch], 0);
}
}
continue;
}
stepSizes[0] = __max (rms[0], BA_EPS);
for (b = 1; b < __min (LF, maxSfbInCh); b++) // apply steeper low-frequency simultaneous masking slopes
{
maskingSlope = (stepSizes[b - 1] + (msOffset << (9u - b))) >> (msShift + 9u - b);
stepSizes[b] = __max (rms[b], maskingSlope + BA_EPS);
}
for (/*b*/; b < __min (MF, maxSfbInCh); b++) // apply typical mid-frequency simultaneous masking slopes
{
maskingSlope = (stepSizes[b - 1] + msOffset) >> msShift;
stepSizes[b] = __max (rms[b], maskingSlope + BA_EPS);
}
if ((samplingRate >= 27713) && (samplingRate < 75132))
{
for (/*b*/; b < __min (HF, maxSfbInCh); b++) // compensate high-frequency slopes for linear SFB width
{
maskingSlope = ((uint64_t) stepSizes[b - 1] * (9u + b - MF) + (msOffset << 3u)) >> (msShift + 3u);
stepSizes[b] = __max (rms[b], maskingSlope + BA_EPS);
}
for (/*b = HF region*/; b < maxSfbInCh; b++) // apply extra high-frequency equal-loudness attenuation
{
for (unsigned d = b - HF; d > 0; d--)
{
elw = (elw * 52430 - SHRT_MIN) >> 16; // elw *= 4/5
}
rmsEqualLoud = uint32_t (((uint64_t) rms[b] * elw - SHRT_MIN) >> 16); // equal loudness weighting
maskingSlope = ((uint64_t) stepSizes[b - 1] * (9u + b - MF) + (msOffset << 3u)) >> (msShift + 3u);
stepSizes[b] = __max (rmsEqualLoud, maskingSlope + BA_EPS);
}
}
else // no equal-loudness weighting for low or high rates
{
for (/*b = MF region*/; b < maxSfbInCh; b++) // compensate high-frequency slopes for linear SFB width
{
maskingSlope = ((uint64_t) stepSizes[b - 1] * (9u + b - MF) + (msOffset << 3u)) >> (msShift + 3u);
stepSizes[b] = __max (rms[b], maskingSlope + BA_EPS);
}
}
stepSizes[b] = 0;
for (b -= 1; b > __min (MF, maxSfbInCh); b--) // complete simultaneous masking by reversing the pattern
{
sumStepSizes += intSqrt (stepSizes[b]);
maskingSlope = ((uint64_t) stepSizes[b] * (8u + b - MF) + (msOffset << 3u)) >> (msShift + 3u);
stepSizes[b - 1] = __max (stepSizes[b - 1], maskingSlope);
}
for (/*b*/; b > __min (LF, maxSfbInCh); b--) // typical reversed mid-freq. simultaneous masking slopes
{
sumStepSizes += intSqrt (stepSizes[b]);
maskingSlope = (stepSizes[b] + msOffset) >> msShift;
stepSizes[b - 1] = __max (stepSizes[b - 1], maskingSlope);
}
for (/*b = min (9, maxSfbInCh)*/; b > 0; b--) // steeper reversed low-freq. simultaneous masking slopes
{
sumStepSizes += intSqrt (stepSizes[b]);
maskingSlope = (stepSizes[b] + (msOffset << (10u - b))) >> (msShift + 10u - b);
stepSizes[b - 1] = __max (stepSizes[b - 1], maskingSlope);
}
sumStepSizes += intSqrt (stepSizes[0]);
// --- LONG window: apply perceptual JND model and local band-peak smoothing, undo equal-loudness weighting
nMeans++;
m_avgStepSize[ch] = __min (USHRT_MAX, (sumStepSizes + (nBandsInCh >> 1)) / nBandsInCh);
sumMeans += m_avgStepSize[ch];
m_avgStepSize[ch] *= m_avgStepSize[ch];
jndPowerLawAndPeakSmoothing (stepSizes, maxSfbInCh, (m_rateIndex == 0) && (samplingRate >= 27713),
m_avgStepSize[ch], m_avgSpecFlat[ch], tnsDisabled ? m_avgTempFlat[ch] : 0);
if ((samplingRate >= 27713) && (samplingRate < 75132))
{
elw = 36; // 36/32 = 9/8
for (b = HF; b < maxSfbInCh; b++) // undo above additional high-frequency equal-loudness attenuation
{
for (unsigned d = b - HF; d > 0; d--)
{
elw = (16u + elw * 40) >> 5; // inverse elw *= 5/4. NOTE: this may overflow for 64 kHz, that's OK
}
if (elw == 138 || elw >= 1024) elw--;
rmsEqualLoud = uint32_t (__min (UINT_MAX, (16u + (uint64_t) stepSizes[b] * elw) >> 5));
stepSizes[b] = rmsEqualLoud;
}
}
} // for ch
if ((nMeans < 2) || (sumMeans <= nMeans * BA_EPS)) // in case of one channel or low-RMS input, we're done
{
return 0; // no error
}
sumMeans = (sumMeans + (nMeans >> 1)) / nMeans;
sumMeans *= sumMeans; // since we've averaged square-roots
for (unsigned ch = 0; ch < nChannels; ch++)
{
const SfbGroupData& grpData = *groupData[ch];
const uint32_t maxSfbInCh = __min (MAX_NUM_SWB_LONG, grpData.sfbsPerGroup);
const uint32_t nBandsInCh = grpData.numWindowGroups * maxSfbInCh;
const uint32_t chStepSize = m_avgStepSize[ch];
uint32_t* stepSizes = &sfbStepSizes[ch * numSwbShort * NUM_WINDOW_GROUPS];
// --- apply INTER-channel simultaneous masking and JND modeling to calculated INTRA-channel step-size data
uint64_t mAvgStepSize; // modified and averaged step-size
if ((nBandsInCh == 0) || (grpData.numWindowGroups > NUM_WINDOW_GROUPS) || (ch == lfeChannelIndex))
{
continue;
}
mAvgStepSize = jndModel (chStepSize, sumMeans, 7 << 6 /*7/8*/, 512);
for (unsigned gr = 0; gr < grpData.numWindowGroups; gr++)
{
uint32_t* gStepSizes = &stepSizes[numSwbShort * gr];
for (unsigned b = 0; b < maxSfbInCh; b++)
{
gStepSizes[b] = uint32_t (__min (UINT_MAX, (mAvgStepSize * gStepSizes[b] + (chStepSize >> 1)) / chStepSize));
}
} // for gr
m_avgStepSize[ch] = (uint32_t) mAvgStepSize;
} // for ch
return 0; // no error
}
unsigned BitAllocator::imprSfbStepSizes (const SfbGroupData* const groupData[USAC_MAX_NUM_CHANNELS], const uint8_t numSwbShort,
const int32_t* const mdctSpec[USAC_MAX_NUM_CHANNELS], const unsigned nSamplesInFrame,
const unsigned nChannels, const unsigned samplingRate, uint32_t* const sfbStepSizes,
const unsigned firstChannelIndex, const uint8_t* const sfm, const bool commonWindow,
const uint8_t* const sfbStereoData /*= nullptr*/, const uint8_t stereoConfig /*= 0*/)
{
const uint8_t maxSfbL16k = 16 + __min (4 + (samplingRate >> 10), (9 << 17) / __max (1, samplingRate)); // SFB index at 15.8 kHz
const uint32_t redFactor = __max ((samplingRate < 25495 ? 2 : 1), __min (3, m_rateIndex)) - (stereoConfig >> 3);
const uint32_t redWeight = __min (4, 9 - __min (9, m_rateIndex));
short* const tempCoeffs = (short* const) m_tempSfbValue;
if ((groupData == nullptr) || (mdctSpec == nullptr) || (sfbStepSizes == nullptr) || (sfm == nullptr) || (nSamplesInFrame > 2048) ||
(numSwbShort < MIN_NUM_SWB_SHORT) || (numSwbShort > MAX_NUM_SWB_SHORT) || (nChannels > USAC_MAX_NUM_CHANNELS) ||
(samplingRate < 7350) || (samplingRate > 96000) || (firstChannelIndex > USAC_MAX_NUM_CHANNELS))
{
return 1; // invalid arguments error
}
for (unsigned ch = firstChannelIndex; ch < firstChannelIndex + nChannels; ch++)
{
const SfbGroupData& grpData = *groupData[ch];
const uint32_t maxSfbInCh = __min (MAX_NUM_SWB_LONG, grpData.sfbsPerGroup);
const bool eightShorts = (grpData.numWindowGroups != 1);
const bool lowRateTuning = (m_rateIndex == 0) && (samplingRate >= 25495 && sfm[ch] <= (SCHAR_MAX >> 1));
const bool undercodingRed = (m_rateIndex > 0) || (samplingRate >= 25495 && sfm[ch] * 8 > UCHAR_MAX * 7) || lowRateTuning;
const uint32_t* rms = grpData.sfbRmsValues;
uint32_t* stepSizes = &sfbStepSizes[ch * numSwbShort * NUM_WINDOW_GROUPS];
if ((grpData.numWindowGroups * maxSfbInCh == 0) || (grpData.numWindowGroups > NUM_WINDOW_GROUPS))
{
continue;
}
for (unsigned gr = 0; gr < grpData.numWindowGroups; gr++)
{
const uint16_t* grpOff = &grpData.sfbOffsets[numSwbShort * gr];
const uint8_t* grpSte = (sfbStereoData == nullptr ? nullptr : &sfbStereoData[numSwbShort * gr]);
const uint32_t* grpRms = &rms[numSwbShort * gr];
const uint32_t* refRms = &groupData[firstChannelIndex + nChannels - 1 - ch]->sfbRmsValues[numSwbShort * gr];
uint32_t* grpStepSizes = &stepSizes[numSwbShort * gr];
uint32_t b, grpRmsMin = INT_MAX; // min. RMS value, used for overcoding reduction
uint64_t s = (eightShorts ? (nSamplesInFrame * grpData.windowGroupLength[gr]) >> 1 : nSamplesInFrame << 2);
memset (m_tempSfbValue, UCHAR_MAX, maxSfbInCh * sizeof (uint8_t));
if (lowRateTuning && (maxSfbInCh > 0) && !eightShorts)
{
uint32_t numRedBands = nSamplesInFrame; // final result lies between 1/4 and 1/2
if ((nChannels == 2) && commonWindow && (grpSte != nullptr))
{
for (b = 0; b < maxSfbInCh; b++) if (grpSte[b] == 0) numRedBands += grpOff[b + 1] - grpOff[b];
}
b = MAX_NUM_SWB_LONG - ((numRedBands * ((SCHAR_MAX >> 1) + 1 - sfm[ch]) + (1 << 11)) >> 12);
while ((b < maxSfbInCh) && (grpRms[b] > grpRms[b - 1])) b++; // start after peak
for (b += ((nChannels == 2) && commonWindow ? b & 1 : 0); b < maxSfbInCh; b++)
{
grpStepSizes[b] = __max (grpStepSizes[b], grpRms[b] >= (UINT_MAX >> 1) ? UINT_MAX : (grpRms[b] + 1) << 1);
}
}
// undercoding reduction for case where large number of coefs is quantized to zero
for (b = 0; b < maxSfbInCh; b++)
{
const uint32_t rmsComp = (grpSte != nullptr && grpSte[b] > 0 ? squareMeanRoot (refRms[b], grpRms[b]) : grpRms[b]);
const uint32_t rmsRef9 = (commonWindow ? refRms[b] >> 9 : rmsComp);
const uint8_t sfbWidth = grpOff[b + 1] - grpOff[b];
if (redWeight > 0 && !eightShorts && sfbWidth > (samplingRate >= 18783 ? 8 : 12)) // transient SFBs
{
const uint32_t gains = m_tnsPredictor->calcParCorCoeffs (&mdctSpec[ch][grpOff[b]], sfbWidth, MAX_PREDICTION_ORDER, tempCoeffs) >> 24;
m_tempSfbValue[b] = UCHAR_MAX - uint8_t ((512u + gains * gains * redWeight) >> (10 + (sfbWidth > 16 ? 0 : (20 - sfbWidth) >> 2)));
if ((b >= 2) && (m_tempSfbValue[b - 1] < m_tempSfbValue[b]) && (m_tempSfbValue[b - 1] < m_tempSfbValue[b - 2]))
{
m_tempSfbValue[b - 1] = __min (m_tempSfbValue[b], m_tempSfbValue[b - 2]); // remove local peaks
}
}
if (grpRms[b] < grpRmsMin) grpRmsMin = grpRms[b];
if (undercodingRed && (rmsComp >= rmsRef9) && (rmsComp < (grpStepSizes[b] >> 1))) // zero-quantized
{
s -= (sfbWidth * redFactor * __min (1u << 11, rmsComp) + (1u << 10)) >> 11;
}
}
if ((samplingRate >= 27713) && (b < maxSfbL16k) && !eightShorts) // zeroed HF data
{
const uint32_t rmsComp = (grpSte != nullptr && grpSte[b] > 0 ? squareMeanRoot (refRms[b], grpRms[b]) : grpRms[b]);
const uint32_t rmsRef9 = (commonWindow ? refRms[b] >> 9 : rmsComp);
const uint8_t sfbWidth = grpOff[maxSfbL16k] - grpOff[b];
if (undercodingRed && (rmsComp >= rmsRef9)) // check only first SFB above max_sfb as simplification
{
s -= (sfbWidth * redFactor * __min (1u << 11, rmsComp) + (1u << 10)) >> 11;
}
}
s = (eightShorts ? s / ((nSamplesInFrame * grpData.windowGroupLength[gr]) >> 8) : s / (nSamplesInFrame >> 5));
if (redWeight > 0 && !eightShorts) memset (tempCoeffs /*= m_tempSfbValue*/, UCHAR_MAX, MAX_PREDICTION_ORDER * sizeof (short));
for (b = 0; b < maxSfbInCh; b++) // improve step-sizes by limiting and attenuation
{
grpStepSizes[b] = uint32_t ((__max (grpRmsMin, grpStepSizes[b]) * s * (m_tempSfbValue[b] + 1u) + (1u << 14)) >> 15);
if (grpStepSizes[b] <= (grpRms[b] >> 11)) grpStepSizes[b] = __max (BA_EPS, grpRms[b] >> 11);
if (lowRateTuning) // clip near-0 SNRs to minimum SNR
{
const uint32_t lim = uint32_t ((grpRms[b] * (8192u - (uint64_t) sfm[ch] * sfm[ch]) + (1u << 12)) >> 13);
if ((grpStepSizes[b] > grpRms[b]) && ((grpStepSizes[b] >> 1) <= lim)) grpStepSizes[b] = grpRms[b];
}
}
}
} // for ch
return 0; // no error
}
Reference in New Issue View Git Blame Copy Permalink