comment function Comp_diff_lt_corr() (02afb3ed) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_enc/ivas_stereo_td_analysis.c

+47 −7

Original line number	Diff line number	Diff line
		@@ -173,7 +173,7 @@ int16_t stereo_tdm_ener_analysis(
		NOOP_decision( hCPE, tdm_NOOP_flag_loc, tmp_SM_flag, rms_L, rms_R, &tdm_SM_flag_loc );

		/----------------------------------------------------------------
		* Adjust stereo downmixing adaptation rate factor
		* Adjust stereo down-mixing adaptation rate factor
		* in function of the signal energy. If signal energy is low,
		* adaptation rate factor is lower. This prevent stereo image
		* move on speech offset
		@@ -737,7 +737,7 @@ static int16_t Get_dt_lt_ener(
		/-------------------------------------------------------------------
		* Function NOOP_decision()
		*
		* Set Near Out Of Phase decision
		* Set Near Out Of Phase decision (HW)
		-------------------------------------------------------------------/

		static void NOOP_decision(
		@@ -882,11 +882,19 @@ static float Comp_diff_lt_corr(
		/----------------------------------------------------------------
		* Adjust stereo downmixing adaptation rate factor
		* in function of the signal energy. If signal energy is low,
		* adaptation rate factor is lower. This prevent stereo image
		* move on speech offset
		* adaptation rate factor is lower. This prevents stereo image
		* to move on speech offset.
		* In particular, for input signal that is considered as having low
		* correlation (LRTD), and the left or the right channel has more
		* than the double of long term energy than the other channel,
		* then the maximum long term energy between left and right will
		* be choosen to set up the adaption rate.
		* for all the other cases, the minimum between the 2 channel
		* long term energy will be used
		----------------------------------------------------------------/

		if ( IsSideMono == 0 && hCPE->hStereoClassif->lrtd_mode == 1 && ( ( tdm_lt_rms_R > 2.0f * tdm_lt_rms_L ) \|\| ( tdm_lt_rms_L > 2.0f * tdm_lt_rms_R ) ) )
		if ( IsSideMono == 0 && hCPE->hStereoClassif->lrtd_mode == 1 &&
		( ( tdm_lt_rms_R > 2.0f * tdm_lt_rms_L ) \|\| ( tdm_lt_rms_L > 2.0f * tdm_lt_rms_R ) ) )
		{
		adaprate = M_ADAP * max( tdm_lt_rms_R, tdm_lt_rms_L ) + B_ADAP;
		}
		@@ -898,7 +906,7 @@ static float Comp_diff_lt_corr(
		adaprate = check_bounds( adaprate, MIN_ADAP_FAC, 1.0f );

		/----------------------------------------------------------------
		* In case of unvoiced content (expect when it is part of an onset),
		* In case of unvoiced content (except when it is part of an onset),
		* the adaptation rate is minimal.
		----------------------------------------------------------------/

		@@ -912,13 +920,25 @@ static float Comp_diff_lt_corr(
		d_lt_corr_raw = 0.0f;
		if ( IsSideMono == 0 )
		{
		/----------------------------------------------------------------
		* In case of switching from a different stereo mode or
		* if on of the channel has very low energy and the signal is
		* considered as having a low correlation (LRTD) and the maximum
		* long term energy betweenm the 2 channel is below 10
		* then speed up the adaptation rate
		----------------------------------------------------------------/
		if ( hCPE->last_element_mode != IVAS_CPE_TD \|\| ( ( tdm_lt_rms_R < 1.0f \|\| tdm_lt_rms_L < 1.0f ) && hCPE->hStereoClassif->lrtd_mode == 1 && max( tdm_lt_rms_R, tdm_lt_rms_L ) < 10.0f ) )
		{
		adaprate = 0.98f; /* speed up the adaptation of the long term values to the current values after coming from DFT */
		}
		adaprate_tmp = max( adaprate, 0.8f );
		madaprate = 1.0f - adaprate_tmp; /* madaprate has temporary value, will updated few lines below */

		/----------------------------------------------------------------
		* Compute the short term correlation to mono difference
		* Then compute the instantaneous ratio between L and R channel.
		* The instantaneous ratio is used for bit rate distribution in case of LRTD,
		* while a more convservative ratio is found later and used for downmixing purpose
		----------------------------------------------------------------/
		d_lt_corr_raw = ( adaprate_tmp * corr_LM + madaprate * *tdm_lt_corr_LM ) -
		( adaprate_tmp * corr_RM + madaprate * tdm_lt_corr_RM ); / Short term smooth correlation differences to mono */
		inst_ratio_L = 0.0f;
		@@ -928,6 +948,10 @@ static float Comp_diff_lt_corr(
		inst_ratio_L = 0.667f * inst_ratio_L + 1.0f;
		inst_ratio_L_out = ( 1.0f - cosf( EVS_PI inst_ratio_L / 2.0f ) ) / 2.0f;
		}
		/----------------------------------------------------------------
		* Find a first estimation of the correlation/energy to mono
		* of L and R difference
		----------------------------------------------------------------/
		madaprate = 1.0f - adaprate;
		corr_RM = adaprate * corr_RM + madaprate * *tdm_lt_corr_RM;
		corr_LM = adaprate * corr_LM + madaprate * *tdm_lt_corr_LM;
		@@ -940,6 +964,22 @@ static float Comp_diff_lt_corr(

		/----------------------------------------------------------------
		* Correct the estimation depending of channels energies evolution
		* In case of out of phase signal and the rms of both L and R channels
		* are low and the long term rms of each channel is below 80% of
		* its current rms value and the long term rms of one of the channel
		* is more than the double of the other channel
		* then,
		* the long term corr evolved slowly using 80% of its long term
		* value and 20% of the current correlation. it is further
		* amplified by a factor of 2.5
		*
		* else if the energy differential of both channel are going in the
		* same direction and the variation of the correlation diffirence
		* is low and the energy of at least one channel is not too low
		* then,
		* the long term corr uses the values estimated above
		* else
		* the long term corr average its values with the current correlation
		----------------------------------------------------------------/

		if ( IsSideMono == 1 && ( ( rms_L < RMS_MIN2 && rms_R < RMS_MIN2 ) && ( ( tdm_lt_rms_L < 0.8f * rms_L ) && ( tdm_lt_rms_R < 0.8f * rms_R ) ) &&