From 53cc1a033a3183c9d137b03446db58d2e58c0e0a Mon Sep 17 00:00:00 2001 From: "Timothy B. Terriberry" Date: Fri, 14 Oct 2011 13:38:24 -0700 Subject: [PATCH] Fix the side frame conditional coding rules. b24e5746 introduced changes to LastGainIndex which broke conditional coding for side frames after a mid-only frame (i.e., in a 60 ms frame where the side is coded, not coded, then coded again). These rules were a mess in general, however, because the side channel state kept a different nFramesDecoded count from the mid channel state, and had no way to tell if the prior side frame was coded. This patch attempts to rationalize them by moving the conditional coding decision up to the top level, where all this information is available. The first coded side frame after an uncoded side frame now always uses independent coding. If such a frame is also not the first side frame in an Opus frame, then it doesn't include an LTP scaling parameter (because the LTP state is well-defined). --- silk/dec_API.c | 28 ++++++++++++++++++++++++++-- silk/decode_frame.c | 7 ++++--- silk/decode_indices.c | 18 ++++++------------ silk/decode_parameters.c | 5 +++-- silk/define.h | 5 +++++ silk/enc_API.c | 26 +++++++++++++++++++++++--- silk/encode_indices.c | 18 ++++++------------ silk/fixed/LTP_scale_ctrl_FIX.c | 6 ++++-- silk/fixed/encode_frame_FIX.c | 14 ++++++++------ silk/fixed/find_pred_coefs_FIX.c | 5 +++-- silk/fixed/main_FIX.h | 12 ++++++++---- silk/fixed/process_gains_FIX.c | 3 ++- silk/float/LTP_scale_ctrl_FLP.c | 5 +++-- silk/float/encode_frame_FLP.c | 19 +++++++++++-------- silk/float/find_pred_coefs_FLP.c | 5 +++-- silk/float/main_FLP.h | 12 ++++++++---- silk/float/process_gains_FLP.c | 5 +++-- silk/main.h | 12 ++++++++---- 18 files changed, 134 insertions(+), 71 deletions(-) diff --git a/silk/dec_API.c b/silk/dec_API.c index ade060f9..018173ef 100644 --- a/silk/dec_API.c +++ b/silk/dec_API.c @@ -182,13 +182,21 @@ opus_int silk_Decode( for( n = 0; n < decControl->nChannelsInternal; n++ ) { if( channel_state[ n ].LBRR_flags[ i ] ) { opus_int pulses[ MAX_FRAME_LENGTH ]; + opus_int condCoding; + if( decControl->nChannelsInternal == 2 && n == 0 ) { silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) { silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); } } - silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1 ); + /* Use conditional coding if previous frame available */ + if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) { + condCoding = CODE_CONDITIONALLY; + } else { + condCoding = CODE_INDEPENDENTLY; + } + silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding ); silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType, channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length ); } @@ -230,7 +238,23 @@ opus_int silk_Decode( /* Call decoder for one frame */ for( n = 0; n < decControl->nChannelsInternal; n++ ) { if( n == 0 || ( ( lostFlag != FLAG_PACKET_LOST ? decode_only_middle : psDec->prev_decode_only_middle ) == 0 ) ) { - ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 + delay ], &nSamplesOutDec, lostFlag ); + opus_int FrameIndex; + opus_int condCoding; + + FrameIndex = channel_state[ 0 ].nFramesDecoded - n; + /* Use independent coding if no previous frame available */ + if( FrameIndex <= 0 ) { + condCoding = CODE_INDEPENDENTLY; + } else if( lostFlag == FLAG_DECODE_LBRR ) { + condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY; + } else if( n > 0 && psDec->prev_decode_only_middle ) { + /* If we skipped a side frame in this packet, we don't + need LTP scaling; the LTP state is well-defined. */ + condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; + } else { + condCoding = CODE_CONDITIONALLY; + } + ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 + delay ], &nSamplesOutDec, lostFlag, condCoding); } else { silk_memset( &samplesOut1_tmp[ n ][ 2 + delay ], 0, nSamplesOutDec * sizeof( opus_int16 ) ); } diff --git a/silk/decode_frame.c b/silk/decode_frame.c index 24786174..54bb920b 100644 --- a/silk/decode_frame.c +++ b/silk/decode_frame.c @@ -40,7 +40,8 @@ opus_int silk_decode_frame( ec_dec *psRangeDec, /* I/O Compressor data structure */ opus_int16 pOut[], /* O Pointer to output speech frame */ opus_int32 *pN, /* O Pointer to size of output frame */ - opus_int lostFlag /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int condCoding /* I The type of conditional coding to use */ ) { silk_decoder_control sDecCtrl; @@ -62,7 +63,7 @@ TIC(DECODE_FRAME) /* Decode quantization indices of side info */ /*********************************************/ TIC(decode_indices) - silk_decode_indices( psDec, psRangeDec, psDec->nFramesDecoded, lostFlag ); + silk_decode_indices( psDec, psRangeDec, psDec->nFramesDecoded, lostFlag, condCoding ); TOC(decode_indices) /*********************************************/ @@ -77,7 +78,7 @@ TOC(decode_pulses) /* Decode parameters and pulse signal */ /********************************************/ TIC(decode_params) - silk_decode_parameters( psDec, &sDecCtrl ); + silk_decode_parameters( psDec, &sDecCtrl, condCoding ); TOC(decode_params) /* Update length. Sampling frequency may have changed */ diff --git a/silk/decode_indices.c b/silk/decode_indices.c index 9ea0a80e..8ac53f4b 100644 --- a/silk/decode_indices.c +++ b/silk/decode_indices.c @@ -36,21 +36,15 @@ void silk_decode_indices( silk_decoder_state *psDec, /* I/O State */ ec_dec *psRangeDec, /* I/O Compressor data structure */ opus_int FrameIndex, /* I Frame number */ - opus_int decode_LBRR /* I Flag indicating LBRR data is being decoded */ + opus_int decode_LBRR, /* I Flag indicating LBRR data is being decoded */ + opus_int condCoding /* I The type of conditional coding to use */ ) { - opus_int i, k, Ix, condCoding; + opus_int i, k, Ix; opus_int decode_absolute_lagIndex, delta_lagIndex; opus_int16 ec_ix[ MAX_LPC_ORDER ]; opus_uint8 pred_Q8[ MAX_LPC_ORDER ]; - /* Use conditional coding if previous frame available */ - if( FrameIndex > 0 && ( decode_LBRR == 0 || psDec->LBRR_flags[ FrameIndex - 1 ] == 1 ) ) { - condCoding = 1; - } else { - condCoding = 0; - } - /*******************************************/ /* Decode signal type and quantizer offset */ /*******************************************/ @@ -66,7 +60,7 @@ void silk_decode_indices( /* Decode gains */ /****************/ /* First subframe */ - if( condCoding ) { + if( condCoding == CODE_CONDITIONALLY ) { /* Conditional coding */ psDec->indices.GainsIndices[ 0 ] = (opus_int8)ec_dec_icdf( psRangeDec, silk_delta_gain_iCDF, 8 ); } else { @@ -110,7 +104,7 @@ void silk_decode_indices( /*********************/ /* Get lag index */ decode_absolute_lagIndex = 1; - if( condCoding && psDec->ec_prevSignalType == TYPE_VOICED ) { + if( condCoding == CODE_CONDITIONALLY && psDec->ec_prevSignalType == TYPE_VOICED ) { /* Decode Delta index */ delta_lagIndex = (opus_int16)ec_dec_icdf( psRangeDec, silk_pitch_delta_iCDF, 8 ); if( delta_lagIndex > 0 ) { @@ -142,7 +136,7 @@ void silk_decode_indices( /**********************/ /* Decode LTP scaling */ /**********************/ - if( !condCoding ) { + if( condCoding == CODE_INDEPENDENTLY ) { psDec->indices.LTP_scaleIndex = (opus_int8)ec_dec_icdf( psRangeDec, silk_LTPscale_iCDF, 8 ); } else { psDec->indices.LTP_scaleIndex = 0; diff --git a/silk/decode_parameters.c b/silk/decode_parameters.c index ad4788f9..1ab8ed74 100644 --- a/silk/decode_parameters.c +++ b/silk/decode_parameters.c @@ -34,7 +34,8 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. /* Decode parameters from payload */ void silk_decode_parameters( silk_decoder_state *psDec, /* I/O State */ - silk_decoder_control *psDecCtrl /* I/O Decoder control */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int condCoding /* I The type of conditional coding to use */ ) { opus_int i, k, Ix; @@ -43,7 +44,7 @@ void silk_decode_parameters( /* Dequant Gains */ silk_gains_dequant( psDecCtrl->Gains_Q16, psDec->indices.GainsIndices, - &psDec->LastGainIndex, psDec->nFramesDecoded, psDec->nb_subfr ); + &psDec->LastGainIndex, condCoding == CODE_CONDITIONALLY, psDec->nb_subfr ); /****************/ /* Decode NLSFs */ diff --git a/silk/define.h b/silk/define.h index f56ca2d7..8f9935fc 100644 --- a/silk/define.h +++ b/silk/define.h @@ -66,6 +66,11 @@ extern "C" #define TYPE_UNVOICED 1 #define TYPE_VOICED 2 +/* Conditional coding types */ +#define CODE_INDEPENDENTLY 0 +#define CODE_INDEPENDENTLY_NO_LTP_SCALING 1 +#define CODE_CONDITIONALLY 2 + /* Settings for stereo processing */ #define STEREO_QUANT_TAB_SIZE 16 #define STEREO_QUANT_SUB_STEPS 5 diff --git a/silk/enc_API.c b/silk/enc_API.c index 339dafc4..7177d4d5 100644 --- a/silk/enc_API.c +++ b/silk/enc_API.c @@ -330,6 +330,8 @@ opus_int silk_Encode( for( i = 0; i < psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket; i++ ) { for( n = 0; n < encControl->nChannelsInternal; n++ ) { if( psEnc->state_Fxx[ n ].sCmn.LBRR_flags[ i ] ) { + opus_int condCoding; + if( encControl->nChannelsInternal == 2 && n == 0 ) { silk_stereo_encode_pred( psRangeEnc, psEnc->sStereo.predIx[ i ] ); /* For LBRR data there's no need to code the mid-only flag if the side-channel LBRR flag is set */ @@ -337,7 +339,13 @@ opus_int silk_Encode( silk_stereo_encode_mid_only( psRangeEnc, psEnc->sStereo.mid_only_flags[ i ] ); } } - silk_encode_indices( &psEnc->state_Fxx[ n ].sCmn, psRangeEnc, i, 1 ); + /* Use conditional coding if previous frame available */ + if( i > 0 && psEnc->state_Fxx[ n ].sCmn.LBRR_flags[ i - 1 ] ) { + condCoding = CODE_CONDITIONALLY; + } else { + condCoding = CODE_INDEPENDENTLY; + } + silk_encode_indices( &psEnc->state_Fxx[ n ].sCmn, psRangeEnc, i, 1, condCoding ); silk_encode_pulses( psRangeEnc, psEnc->state_Fxx[ n ].sCmn.indices_LBRR[i].signalType, psEnc->state_Fxx[ n ].sCmn.indices_LBRR[i].quantOffsetType, psEnc->state_Fxx[ n ].sCmn.pulses_LBRR[ i ], psEnc->state_Fxx[ n ].sCmn.frame_length ); } @@ -400,7 +408,6 @@ opus_int silk_Encode( psEnc->state_Fxx[ 1 ].sCmn.prevSignalType = TYPE_NO_VOICE_ACTIVITY; psEnc->state_Fxx[ 1 ].sCmn.sNSQ.prev_inv_gain_Q16 = 65536; } - psEnc->prev_decode_only_middle = psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded ]; /* Encode */ for( n = 0; n < encControl->nChannelsInternal; n++ ) { @@ -411,9 +418,21 @@ opus_int silk_Encode( } if( channelRate_bps > 0 ) { + opus_int condCoding; + silk_control_SNR( &psEnc->state_Fxx[ n ].sCmn, channelRate_bps ); - if( ( ret = silk_encode_frame_Fxx( &psEnc->state_Fxx[ n ], nBytesOut, psRangeEnc ) ) != 0 ) { + /* Use independent coding if no previous frame available */ + if( psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded - n <= 0 ) { + condCoding = CODE_INDEPENDENTLY; + } else if( n > 0 && psEnc->prev_decode_only_middle ) { + /* If we skipped a side frame in this packet, we don't + need LTP scaling; the LTP state is well-defined. */ + condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; + } else { + condCoding = CODE_CONDITIONALLY; + } + if( ( ret = silk_encode_frame_Fxx( &psEnc->state_Fxx[ n ], nBytesOut, psRangeEnc, condCoding ) ) != 0 ) { silk_assert( 0 ); } psEnc->state_Fxx[ n ].sCmn.nFramesEncoded++; @@ -421,6 +440,7 @@ opus_int silk_Encode( psEnc->state_Fxx[ n ].sCmn.controlled_since_last_payload = 0; psEnc->state_Fxx[ n ].sCmn.inputBufIx = 0; } + psEnc->prev_decode_only_middle = psEnc->sStereo.mid_only_flags[ psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded - 1 ]; /* Insert VAD and FEC flags at beginning of bitstream */ if( *nBytesOut > 0 && psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded == psEnc->state_Fxx[ 0 ].sCmn.nFramesPerPacket) { diff --git a/silk/encode_indices.c b/silk/encode_indices.c index 6d8e82a1..f87de036 100644 --- a/silk/encode_indices.c +++ b/silk/encode_indices.c @@ -36,10 +36,11 @@ void silk_encode_indices( silk_encoder_state *psEncC, /* I/O Encoder state */ ec_enc *psRangeEnc, /* I/O Compressor data structure */ opus_int FrameIndex, /* I Frame number */ - opus_int encode_LBRR /* I Flag indicating LBRR data is being encoded */ + opus_int encode_LBRR, /* I Flag indicating LBRR data is being encoded */ + opus_int condCoding /* I The type of conditional coding to use */ ) { - opus_int i, k, condCoding, typeOffset; + opus_int i, k, typeOffset; opus_int encode_absolute_lagIndex, delta_lagIndex; opus_int16 ec_ix[ MAX_LPC_ORDER ]; opus_uint8 pred_Q8[ MAX_LPC_ORDER ]; @@ -49,13 +50,6 @@ void silk_encode_indices( opus_int nBytes_after, nBytes_before; #endif - /* Use conditional coding if previous frame available */ - if( FrameIndex > 0 && ( encode_LBRR == 0 || psEncC->LBRR_flags[ FrameIndex - 1 ] == 1 ) ) { - condCoding = 1; - } else { - condCoding = 0; - } - if( encode_LBRR ) { psIndices = &psEncC->indices_LBRR[ FrameIndex ]; } else { @@ -81,7 +75,7 @@ void silk_encode_indices( nBytes_before = silk_RSHIFT( ec_tell( psRangeEnc ) + 7, 3 ); #endif /* first subframe */ - if( condCoding ) { + if( condCoding == CODE_CONDITIONALLY ) { /* conditional coding */ silk_assert( psIndices->GainsIndices[ 0 ] >= 0 && psIndices->GainsIndices[ 0 ] < MAX_DELTA_GAIN_QUANT - MIN_DELTA_GAIN_QUANT + 1 ); ec_enc_icdf( psRangeEnc, psIndices->GainsIndices[ 0 ], silk_delta_gain_iCDF, 8 ); @@ -148,7 +142,7 @@ void silk_encode_indices( #endif /* lag index */ encode_absolute_lagIndex = 1; - if( condCoding && psEncC->ec_prevSignalType == TYPE_VOICED ) { + if( condCoding == CODE_CONDITIONALLY && psEncC->ec_prevSignalType == TYPE_VOICED ) { /* Delta Encoding */ delta_lagIndex = psIndices->lagIndex - psEncC->ec_prevLagIndex; if( delta_lagIndex < -8 || delta_lagIndex > 11 ) { @@ -212,7 +206,7 @@ void silk_encode_indices( /**********************/ /* Encode LTP scaling */ /**********************/ - if( !condCoding ) { + if( condCoding == CODE_INDEPENDENTLY ) { silk_assert( psIndices->LTP_scaleIndex >= 0 && psIndices->LTP_scaleIndex < 3 ); ec_enc_icdf( psRangeEnc, psIndices->LTP_scaleIndex, silk_LTPscale_iCDF, 8 ); } diff --git a/silk/fixed/LTP_scale_ctrl_FIX.c b/silk/fixed/LTP_scale_ctrl_FIX.c index 463cf735..5e85cc74 100644 --- a/silk/fixed/LTP_scale_ctrl_FIX.c +++ b/silk/fixed/LTP_scale_ctrl_FIX.c @@ -33,7 +33,9 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. void silk_LTP_scale_ctrl_FIX( silk_encoder_state_FIX *psEnc, /* I/O encoder state FIX */ - silk_encoder_control_FIX *psEncCtrl /* I/O encoder control FIX */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control FIX */ + opus_int condCoding /* I The type of conditional coding to use */ + ) { opus_int round_loss; @@ -44,7 +46,7 @@ void silk_LTP_scale_ctrl_FIX( psEnc->prevLTPredCodGain_Q7 = psEncCtrl->LTPredCodGain_Q7; /* Only scale if first frame in packet */ - if( psEnc->sCmn.nFramesEncoded == 0 ) { + if( condCoding == CODE_INDEPENDENTLY ) { round_loss = psEnc->sCmn.PacketLoss_perc + psEnc->sCmn.nFramesPerPacket - 1; psEnc->sCmn.indices.LTP_scaleIndex = (opus_int8)silk_LIMIT( silk_SMULWB( silk_SMULBB( round_loss, psEnc->HPLTPredCodGain_Q7 ), SILK_FIX_CONST( 0.1, 9 ) ), 0, 2 ); diff --git a/silk/fixed/encode_frame_FIX.c b/silk/fixed/encode_frame_FIX.c index 6f4367f7..c3da9bcf 100644 --- a/silk/fixed/encode_frame_FIX.c +++ b/silk/fixed/encode_frame_FIX.c @@ -38,7 +38,8 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. opus_int silk_encode_frame_FIX( silk_encoder_state_FIX *psEnc, /* I/O Encoder state FIX */ opus_int32 *pnBytesOut, /* O Number of payload bytes */ - ec_enc *psRangeEnc /* I/O compressor data structure */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding /* I The type of conditional coding to use */ ) { silk_encoder_control_FIX sEncCtrl; @@ -116,7 +117,7 @@ TOC(NOISE_SHAPE_ANALYSIS) /* Find linear prediction coefficients (LPC + LTP) */ /***************************************************/ TIC(FIND_PRED_COEF) - silk_find_pred_coefs_FIX( psEnc, &sEncCtrl, res_pitch, x_frame ); + silk_find_pred_coefs_FIX( psEnc, &sEncCtrl, res_pitch, x_frame, condCoding ); TOC(FIND_PRED_COEF) /****************************************/ @@ -137,7 +138,7 @@ TOC(PREFILTER) /* Low Bitrate Redundant Encoding */ /****************************************/ TIC(LBRR) - silk_LBRR_encode_FIX( psEnc, &sEncCtrl, xfw ); + silk_LBRR_encode_FIX( psEnc, &sEncCtrl, xfw, condCoding ); TOC(LBRR) /*****************************************/ @@ -174,7 +175,7 @@ TOC(NSQ) /* Encode Parameters */ /****************************************/ TIC(ENCODE_PARAMS) - silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0 ); + silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding ); TOC(ENCODE_PARAMS) /****************************************/ @@ -242,7 +243,8 @@ TOC(ENCODE_FRAME) void silk_LBRR_encode_FIX( silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ silk_encoder_control_FIX *psEncCtrl, /* I/O Pointer to Silk FIX encoder control struct */ - const opus_int16 xfw[] /* I Input signal */ + const opus_int16 xfw[], /* I Input signal */ + opus_int condCoding /* I The type of conditional coding used so far for this frame */ ) { opus_int32 TempGains_Q16[ MAX_NB_SUBFR ]; @@ -274,7 +276,7 @@ void silk_LBRR_encode_FIX( /* Decode to get gains in sync with decoder */ /* Overwrite unquantized gains with quantized gains */ silk_gains_dequant( psEncCtrl->Gains_Q16, psIndices_LBRR->GainsIndices, - &psEnc->sCmn.LBRRprevLastGainIndex, psEnc->sCmn.nFramesEncoded, psEnc->sCmn.nb_subfr ); + &psEnc->sCmn.LBRRprevLastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); /*****************************************/ /* Noise shaping quantization */ diff --git a/silk/fixed/find_pred_coefs_FIX.c b/silk/fixed/find_pred_coefs_FIX.c index 980e321d..e9891cd4 100644 --- a/silk/fixed/find_pred_coefs_FIX.c +++ b/silk/fixed/find_pred_coefs_FIX.c @@ -35,7 +35,8 @@ void silk_find_pred_coefs_FIX( silk_encoder_state_FIX *psEnc, /* I/O encoder state */ silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ const opus_int16 res_pitch[], /* I Residual from pitch analysis */ - const opus_int16 x[] /* I Speech signal */ + const opus_int16 x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ ) { opus_int i; @@ -86,7 +87,7 @@ void silk_find_pred_coefs_FIX( WLTP, psEnc->sCmn.mu_LTP_Q9, psEnc->sCmn.LTPQuantLowComplexity, psEnc->sCmn.nb_subfr); /* Control LTP scaling */ - silk_LTP_scale_ctrl_FIX( psEnc, psEncCtrl ); + silk_LTP_scale_ctrl_FIX( psEnc, psEncCtrl, condCoding ); /* Create LTP residual */ silk_LTP_analysis_filter_FIX( LPC_in_pre, x - psEnc->sCmn.predictLPCOrder, psEncCtrl->LTPCoef_Q14, diff --git a/silk/fixed/main_FIX.h b/silk/fixed/main_FIX.h index e0b1e054..56ba1ef0 100644 --- a/silk/fixed/main_FIX.h +++ b/silk/fixed/main_FIX.h @@ -59,7 +59,8 @@ void silk_HP_variable_cutoff( opus_int silk_encode_frame_FIX( silk_encoder_state_FIX *psEnc, /* I/O Pointer to Silk FIX encoder state */ opus_int32 *pnBytesOut, /* O Pointer to number of payload bytes; */ - ec_enc *psRangeEnc /* I/O compressor data structure */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding /* I The type of conditional coding to use */ ); /* Low Bitrate Redundancy (LBRR) encoding. Reuse all parameters but encode with lower bitrate */ @@ -118,7 +119,8 @@ void silk_warped_autocorrelation_FIX( /* Calculation of LTP state scaling */ void silk_LTP_scale_ctrl_FIX( silk_encoder_state_FIX *psEnc, /* I/O encoder state */ - silk_encoder_control_FIX *psEncCtrl /* I/O encoder control */ + silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ + opus_int condCoding /* I The type of conditional coding to use */ ); /**********************************************/ @@ -137,7 +139,8 @@ void silk_find_pred_coefs_FIX( silk_encoder_state_FIX *psEnc, /* I/O encoder state */ silk_encoder_control_FIX *psEncCtrl, /* I/O encoder control */ const opus_int16 res_pitch[], /* I Residual from pitch analysis */ - const opus_int16 x[] /* I Speech signal */ + const opus_int16 x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ ); /* LPC analysis */ @@ -204,7 +207,8 @@ opus_int32 silk_residual_energy16_covar_FIX( /* Processing of gains */ void silk_process_gains_FIX( silk_encoder_state_FIX *psEnc, /* I/O Encoder state */ - silk_encoder_control_FIX *psEncCtrl /* I/O Encoder control */ + silk_encoder_control_FIX *psEncCtrl, /* I/O Encoder control */ + opus_int condCoding /* I The type of conditional coding to use */ ); /******************/ diff --git a/silk/fixed/process_gains_FIX.c b/silk/fixed/process_gains_FIX.c index 0dea69b6..4fe55bb0 100644 --- a/silk/fixed/process_gains_FIX.c +++ b/silk/fixed/process_gains_FIX.c @@ -36,6 +36,7 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. void silk_process_gains_FIX( silk_encoder_state_FIX *psEnc, /* I/O Encoder state_FIX */ silk_encoder_control_FIX *psEncCtrl /* I/O Encoder control_FIX */ + opus_int condCoding /* The type of conditional coding to use */ ) { silk_shape_state_FIX *psShapeSt = &psEnc->sShape; @@ -87,7 +88,7 @@ void silk_process_gains_FIX( /* Noise shaping quantization */ silk_gains_quant( psEnc->sCmn.indices.GainsIndices, psEncCtrl->Gains_Q16, - &psShapeSt->LastGainIndex, psEnc->sCmn.nFramesEncoded, psEnc->sCmn.nb_subfr ); + &psShapeSt->LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); /* Set quantizer offset for voiced signals. Larger offset when LTP coding gain is low or tilt is high (ie low-pass) */ if( psEnc->sCmn.indices.signalType == TYPE_VOICED ) { diff --git a/silk/float/LTP_scale_ctrl_FLP.c b/silk/float/LTP_scale_ctrl_FLP.c index 76290b14..cbac1005 100644 --- a/silk/float/LTP_scale_ctrl_FLP.c +++ b/silk/float/LTP_scale_ctrl_FLP.c @@ -33,7 +33,8 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. void silk_LTP_scale_ctrl_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ - silk_encoder_control_FLP *psEncCtrl /* I/O Encoder control FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ ) { opus_int round_loss; @@ -45,7 +46,7 @@ void silk_LTP_scale_ctrl_FLP( psEnc->prevLTPredCodGain = psEncCtrl->LTPredCodGain; /* Only scale if first frame in packet */ - if( psEnc->sCmn.nFramesEncoded == 0 ) { + if( condCoding == CODE_INDEPENDENTLY ) { round_loss = psEnc->sCmn.PacketLoss_perc + psEnc->sCmn.nFramesPerPacket; psEnc->sCmn.indices.LTP_scaleIndex = (opus_int8)silk_LIMIT( round_loss * psEnc->HPLTPredCodGain * 0.1f, 0.0f, 2.0f ); } else { diff --git a/silk/float/encode_frame_FLP.c b/silk/float/encode_frame_FLP.c index 4d156a03..d1056428 100644 --- a/silk/float/encode_frame_FLP.c +++ b/silk/float/encode_frame_FLP.c @@ -36,7 +36,8 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. static inline void silk_LBRR_encode_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ - const silk_float xfw[] /* I Input signal */ + const silk_float xfw[], /* I Input signal */ + opus_int condCoding /* I The type of conditional coding used so far for this frame */ ); /****************/ @@ -45,7 +46,8 @@ static inline void silk_LBRR_encode_FLP( opus_int silk_encode_frame_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ opus_int32 *pnBytesOut, /* O Number of payload bytes */ - ec_enc *psRangeEnc /* I/O compressor data structure */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding /* I The type of conditional coding to use */ ) { silk_encoder_control_FLP sEncCtrl; @@ -128,14 +130,14 @@ TOC(NOISE_SHAPE_ANALYSIS) /* Find linear prediction coefficients (LPC + LTP) */ /***************************************************/ TIC(FIND_PRED_COEF) - silk_find_pred_coefs_FLP( psEnc, &sEncCtrl, res_pitch, x_frame ); + silk_find_pred_coefs_FLP( psEnc, &sEncCtrl, res_pitch, x_frame, condCoding ); TOC(FIND_PRED_COEF) /****************************************/ /* Process gains */ /****************************************/ TIC(PROCESS_GAINS) - silk_process_gains_FLP( psEnc, &sEncCtrl ); + silk_process_gains_FLP( psEnc, &sEncCtrl, condCoding ); TOC(PROCESS_GAINS) /*****************************************/ @@ -149,7 +151,7 @@ TOC(PREFILTER) /* Low Bitrate Redundant Encoding */ /****************************************/ TIC(LBRR) - silk_LBRR_encode_FLP( psEnc, &sEncCtrl, xfw ); + silk_LBRR_encode_FLP( psEnc, &sEncCtrl, xfw, condCoding ); TOC(LBRR) /*****************************************/ @@ -178,7 +180,7 @@ TOC(NSQ) /* Encode Parameters */ /****************************************/ TIC(ENCODE_PARAMS) - silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0 ); + silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding ); TOC(ENCODE_PARAMS) /****************************************/ @@ -222,7 +224,8 @@ TOC(ENCODE_FRAME) static inline void silk_LBRR_encode_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ - const silk_float xfw[] /* I Input signal */ + const silk_float xfw[], /* I Input signal */ + opus_int condCoding /* I The type of conditional coding used so far for this frame */ ) { opus_int k; @@ -255,7 +258,7 @@ static inline void silk_LBRR_encode_FLP( /* Decode to get gains in sync with decoder */ silk_gains_dequant( Gains_Q16, psIndices_LBRR->GainsIndices, - &psEnc->sCmn.LBRRprevLastGainIndex, psEnc->sCmn.nFramesEncoded, psEnc->sCmn.nb_subfr ); + &psEnc->sCmn.LBRRprevLastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); /* Overwrite unquantized gains with quantized gains and convert back to Q0 from Q16 */ for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { diff --git a/silk/float/find_pred_coefs_FLP.c b/silk/float/find_pred_coefs_FLP.c index 49c29d36..04881855 100644 --- a/silk/float/find_pred_coefs_FLP.c +++ b/silk/float/find_pred_coefs_FLP.c @@ -36,7 +36,8 @@ void silk_find_pred_coefs_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ const silk_float res_pitch[], /* I Residual from pitch analysis */ - const silk_float x[] /* I Speech signal */ + const silk_float x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ ) { opus_int i; @@ -73,7 +74,7 @@ void silk_find_pred_coefs_FLP( WLTP, psEnc->sCmn.mu_LTP_Q9, psEnc->sCmn.LTPQuantLowComplexity, psEnc->sCmn.nb_subfr ); /* Control LTP scaling */ - silk_LTP_scale_ctrl_FLP( psEnc, psEncCtrl ); + silk_LTP_scale_ctrl_FLP( psEnc, psEncCtrl, condCoding ); /* Create LTP residual */ silk_LTP_analysis_filter_FLP( LPC_in_pre, x - psEnc->sCmn.predictLPCOrder, psEncCtrl->LTPCoef, diff --git a/silk/float/main_FLP.h b/silk/float/main_FLP.h index 605fd384..d0c2de7c 100644 --- a/silk/float/main_FLP.h +++ b/silk/float/main_FLP.h @@ -57,7 +57,8 @@ void silk_HP_variable_cutoff( opus_int silk_encode_frame_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ opus_int32 *pnBytesOut, /* O Number of payload bytes; */ - ec_enc *psRangeEnc /* I/O compressor data structure */ + ec_enc *psRangeEnc, /* I/O compressor data structure */ + opus_int condCoding /* I The type of conditional coding to use */ ); /* Initializes the Silk encoder state */ @@ -108,7 +109,8 @@ void silk_warped_autocorrelation_FLP( /* Calculation of LTP state scaling */ void silk_LTP_scale_ctrl_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ - silk_encoder_control_FLP *psEncCtrl /* I/O Encoder control FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ ); /**********************************************/ @@ -127,7 +129,8 @@ void silk_find_pred_coefs_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ const silk_float res_pitch[], /* I Residual from pitch analysis */ - const silk_float x[] /* I Speech signal */ + const silk_float x[], /* I Speech signal */ + opus_int condCoding /* I The type of conditional coding to use */ ); /* LPC analysis */ @@ -234,7 +237,8 @@ void silk_VQ_WMat_EC_FLP( /* Processing of gains */ void silk_process_gains_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ - silk_encoder_control_FLP *psEncCtrl /* I/O Encoder control FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ ); /******************/ diff --git a/silk/float/process_gains_FLP.c b/silk/float/process_gains_FLP.c index e0c18a97..2c7dfdd0 100644 --- a/silk/float/process_gains_FLP.c +++ b/silk/float/process_gains_FLP.c @@ -35,7 +35,8 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. /* Processing of gains */ void silk_process_gains_FLP( silk_encoder_state_FLP *psEnc, /* I/O Encoder state FLP */ - silk_encoder_control_FLP *psEncCtrl /* I/O Encoder control FLP */ + silk_encoder_control_FLP *psEncCtrl, /* I/O Encoder control FLP */ + opus_int condCoding /* I The type of conditional coding to use */ ) { silk_shape_state_FLP *psShapeSt = &psEnc->sShape; @@ -68,7 +69,7 @@ void silk_process_gains_FLP( /* Noise shaping quantization */ silk_gains_quant( psEnc->sCmn.indices.GainsIndices, pGains_Q16, - &psShapeSt->LastGainIndex, psEnc->sCmn.nFramesEncoded, psEnc->sCmn.nb_subfr ); + &psShapeSt->LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr ); /* Overwrite unquantized gains with quantized gains and convert back to Q0 from Q16 */ for( k = 0; k < psEnc->sCmn.nb_subfr; k++ ) { diff --git a/silk/main.h b/silk/main.h index 85423ab6..1f2f01a7 100644 --- a/silk/main.h +++ b/silk/main.h @@ -377,7 +377,8 @@ opus_int silk_decode_frame( ec_dec *psRangeDec, /* I/O Compressor data structure */ opus_int16 pOut[], /* O Pointer to output speech frame */ opus_int32 *pN, /* O Pointer to size of output frame */ - opus_int lostFlag /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */ + opus_int condCoding /* I The type of conditional coding to use */ ); /* Decode LBRR side info and excitation */ @@ -391,13 +392,15 @@ void silk_decode_indices( silk_decoder_state *psDec, /* I/O State */ ec_dec *psRangeDec, /* I/O Compressor data structure */ opus_int FrameIndex, /* I Frame number */ - opus_int decode_LBRR /* I Flag indicating LBRR data is being decoded */ + opus_int decode_LBRR, /* I Flag indicating LBRR data is being decoded */ + opus_int condCoding /* I The type of conditional coding to use */ ); /* Decode parameters from payload */ void silk_decode_parameters( silk_decoder_state *psDec, /* I/O State */ - silk_decoder_control *psDecCtrl /* I/O Decoder control */ + silk_decoder_control *psDecCtrl, /* I/O Decoder control */ + opus_int condCoding /* I The type of conditional coding to use */ ); /* Core decoder. Performs inverse NSQ operation LTP + LPC */ @@ -439,7 +442,8 @@ void silk_encode_indices( silk_encoder_state *psEncC, /* I/O Encoder state */ ec_enc *psRangeEnc, /* I/O Compressor data structure */ opus_int FrameIndex, /* I Frame number */ - opus_int encode_LBRR /* I Flag indicating LBRR data is being encoded */ + opus_int encode_LBRR, /* I Flag indicating LBRR data is being encoded */ + opus_int condCoding /* I The type of conditional coding to use */ ); #endif