Fix for stereo DMX / PHA mode : Change the resolution and improve the... (bdfb1772) · Commits · SA4 / Audio / IVAS BASOP

lib_enc/ivas_stat_enc.h

+8 −0

Original line number	Diff line number	Diff line
		@@ -1179,7 +1179,11 @@ typedef struct stereo_dmx_evs_correlation_filter_structure
		{
		Word16 init_frmCntr;

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		Word32 isd_rate_s_fx; // Q31
		#else
		Word16 isd_rate_s_fx; // Q15
		#endif
		Word32 iccr_s_fx; // Q31
		Word32 ipd_ff_fx[STEREO_DMX_EVS_NB_SUBBAND_MAX]; // Q31
		Word32 Pr_fx[STEREO_DMX_EVS_NB_SUBBAND_MAX]; // Q31
		@@ -1220,7 +1224,11 @@ typedef struct stereo_dmx_evs_enc_data_structure
		STEREO_DMX_EVS_POC_HANDLE hPOC;
		STEREO_DMX_EVS_PHA_HANDLE hPHA;

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		Word16 itd_fx; // Q0
		#else
		Word32 itd_fx; // Q16
		#endif

		Word32 pre_dmx_energy_fx[1];
		Word16 pre_dmx_energy_fx_e[1];

lib_enc/ivas_stereo_dmx_evs_fx.c

+126 −10

Original line number	Diff line number	Diff line
		@@ -68,11 +68,17 @@
		#define STEREO_DMX_EVS_ISD_THRES_L_Q31 1932735283
		#define STEREO_DMX_EVS_ISD_DIST_THRES_IPD_Q15 ONE_IN_Q14

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		#define STEREO_DMX_EVS_ISD_FORGETTING_Q31 2040109465
		#define STEREO_DMX_EVS_ISD_1MFORGETTING_Q15 1638
		#define STEREO_DMX_EVS_ISD_DIST_HYST_L_Q31 773094113
		#define STEREO_DMX_EVS_ISD_DIST_HYST_H_Q31 923417968
		#define STEREO_DMX_EVS_ISD_INVTHRES_H 1270700383
		#else
		#define STEREO_DMX_EVS_ISD_FORGETTING_Q15 31129
		#define STEREO_DMX_EVS_ISD_DIST_HYST_L_Q15 11796
		#define STEREO_DMX_EVS_ISD_DIST_HYST_H_Q15 14090
		#endif

		#define STEREO_DMX_EVS_ICCR_FORGETTING_Q31 1503238554
		#define STEREO_DMX_EVS_ICCR_HYST_L_Q31 1610612736
		@@ -155,7 +161,11 @@ static ivas_error estimate_itd_fx(
		STEREO_DMX_EVS_PHA_HANDLE hPHA, /* i/o: correlation filter structure */
		const Word32 srcL[], /* i : Lch input signal Q16 */
		const Word32 srcR[], /* i : Rch input signal Q16 */
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		Word16 itd[], /* o : estimated itd Q0 */
		#else
		Word32 itd[], /* o : estimated itd Q16 */
		#endif
		const Word16 input_frame /* i : input frame length per channel */
		);
		static void adapt_gain_fx(
		@@ -198,7 +208,11 @@ static void create_M_signal_fx(
		);
		static Word32 find_poc_peak_fx(
		STEREO_DMX_EVS_POC_HANDLE hPOC, /* i/o: phase only correlation structure */
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		Word16 itd_fx[], /* o : estimated itd */
		#else
		Word32 itd_fx[], /* o : estimated itd */
		#endif
		const Word16 input_frame, /* i : input frame length per channel */
		const Word32 ratio_fixed /* i : adapting ratio */
		);
		@@ -607,12 +621,20 @@ static void calc_poc_fx(
		Dr = L_add( specLr[i], specRr[i] ); // spec_e
		Di = L_add( specLi[i], specRi[i] ); // spec_e
		// if ( ( Nr * Nr + Ni * Ni ) > STEREO_DMX_EVS_ISD_THRES_H * ( Dr * Dr + Di * Di ) )
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		IF( GT_32( Mpy_32_32_r( L_add( Mpy_32_32_r( Nr, Nr ), Mpy_32_32_r( Ni, Ni ) ), STEREO_DMX_EVS_ISD_INVTHRES_H ), L_add( Mpy_32_32_r( Dr, Dr ), Mpy_32_32_r( Di, Di ) ) ) )
		#else
		IF( GT_32( Mpy_32_32_r( L_add( Mpy_32_32_r( Nr, Nr ), Mpy_32_32_r( Ni, Ni ) ), 1270700383 /1/STEREO_DMX_EVS_ISD_THRES_H in Q31/ ), L_add( Mpy_32_32_r( Dr, Dr ), Mpy_32_32_r( Di, Di ) ) ) )
		#endif
		{
		isd_cnt_h = add( isd_cnt_h, 1 );
		}
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		W_tmp = W_add( W_mult0_32_32( Mpy_32_32_r( Dr, STEREO_DMX_EVS_ISD_THRES_L_Q31 ), Dr ), W_mult0_32_32( Mpy_32_32_r( Di, STEREO_DMX_EVS_ISD_THRES_L_Q31 ), Di ) ); // Q62
		IF( LT_64( W_add( W_mult0_32_32( Nr, Nr ), W_mult0_32_32( Ni, Ni ) ), W_tmp ) ) // Q62
		#else
		IF( LT_32( L_add( Mpy_32_32_r( Nr, Nr ), Mpy_32_32_r( Ni, Ni ) ), Mpy_32_32_r( STEREO_DMX_EVS_ISD_THRES_L_Q31, L_add( Mpy_32_32_r( Dr, Dr ), Mpy_32_32_r( Di, Di ) ) ) ) )
		#endif
		{
		isd_cnt_l = add( isd_cnt_l, 1 );
		}
		@@ -621,10 +643,19 @@ static void calc_poc_fx(
		isd_rate = BASOP_Util_Divide1616_Scale( isd_cnt_h, freq_8k, &isd_rate_e );
		// Saturation to handle values close to 1.0f
		isd_rate = shl_sat( isd_rate, isd_rate_e ); // Q15
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		hPHA->isd_rate_s_fx = L_add( Mpy_32_32_r( STEREO_DMX_EVS_ISD_FORGETTING_Q31, hPHA->isd_rate_s_fx ), L_mult( STEREO_DMX_EVS_ISD_1MFORGETTING_Q15, isd_rate ) );
		move32();
		#else
		hPHA->isd_rate_s_fx = add( mult_r( STEREO_DMX_EVS_ISD_FORGETTING_Q15, hPHA->isd_rate_s_fx ), mult_r( MAX_16 - STEREO_DMX_EVS_ISD_FORGETTING_Q15, isd_rate ) );
		move16();
		#endif

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		IF( GT_32( hPHA->isd_rate_s_fx, STEREO_DMX_EVS_ISD_DIST_HYST_H_Q31 ) )
		#else
		IF( GT_16( hPHA->isd_rate_s_fx, STEREO_DMX_EVS_ISD_DIST_HYST_H_Q15 ) )
		#endif
		{
		IF( NE_32( hPHA->curr_pha, STEREO_DMX_EVS_PHA_IPD ) )
		{
		@@ -649,7 +680,11 @@ static void calc_poc_fx(
		hPHA->prev_pha = STEREO_DMX_EVS_PHA_IPD;
		move32();
		}
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		ELSE IF( LT_32( hPHA->isd_rate_s_fx, STEREO_DMX_EVS_ISD_DIST_HYST_L_Q31 ) )
		#else
		ELSE IF( LT_16( hPHA->isd_rate_s_fx, STEREO_DMX_EVS_ISD_DIST_HYST_L_Q15 ) )
		#endif
		{
		IF( NE_32( hPHA->curr_pha, STEREO_DMX_EVS_PHA_IPD2 ) )
		{
		@@ -717,6 +752,7 @@ static void calc_poc_fx(

		/* Energy */

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA

		// Left
		W_tmp = W_add( W_mult0_32_32( specLr[i], specLr[i] ), W_mult0_32_32( specLi[i], specLi[i] ) ); // Q(62-(2specL_e)) -> Q(63 - ((2specL_e) +1))
		@@ -737,6 +773,20 @@ static void calc_poc_fx(
		}
		tEr[n] = BASOP_Util_Add_Mant32Exp( tEr[n], tEr_e[n], W_round64_L( W_tmp ), sub( shl( spec_e, 1 ), sub( L_tmp_e, 1 ) ), &tEr_e[n] );
		move32();
		#else
		tEl[n] = BASOP_Util_Add_Mant32Exp( tEl[n], tEl_e[n], L_add( Mpy_32_32_r( specLr[i], specLr[i] ), Mpy_32_32_r( specLi[i], specLi[i] ) ), shl( spec_e, 1 ), &tEl_e[n] );
		move32();

		// Right
		W_tmp = W_add( W_mult0_32_32( specRr[i], specRr[i] ), W_mult0_32_32( specRi[i], specRi[i] ) ); // Q(62-(2specR_e)) -> Q(63 - ((2specR_e) +1))
		L_tmp_e = W_norm( W_tmp );
		IF( L_tmp_e != 0 )
		{
		W_tmp = W_shl( W_tmp, L_tmp_e ); // Q(63 - ((2*spec_e) - (L_tmp_e - 1))
		}
		tEr[n] = BASOP_Util_Add_Mant32Exp( tEr[n], tEr_e[n], W_round64_L( W_tmp ), sub( shl( spec_e, 1 ), sub( L_tmp_e, 1 ) ), &tEr_e[n] );
		move32();
		#endif

		/* IPD */
		// IPDr = L_add(Mpy_32_32_r(specLr[i], specRr[i]), Mpy_32_32_r(specLi[i], specRi[i])); //2*spec_e
		@@ -767,6 +817,7 @@ static void calc_poc_fx(
		tIPDr = L_sub( Mpy_32_32_r( specRr[i], IPDr ), Mpy_32_32_r( specRi[i], IPDi ) ); // spec_e
		tIPDi = L_add( Mpy_32_32_r( specRr[i], IPDi ), Mpy_32_32_r( specRi[i], IPDr ) ); // spec_e

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA

		Nr = BASOP_Util_Add_Mant32Exp( Nr, Nr_e, L_add( Mpy_32_32_r( specLr[i], tIPDr ), Mpy_32_32_r( specLi[i], tIPDi ) ), 0, &Nr_e );
		Ni = BASOP_Util_Add_Mant32Exp( Ni, Ni_e, L_sub( Mpy_32_32_r( specLi[i], tIPDr ), Mpy_32_32_r( specLr[i], tIPDi ) ), 0, &Ni_e );
		@@ -788,6 +839,13 @@ static void calc_poc_fx(
		W_tmp = W_shl( W_tmp, L_tmp_e );
		}
		eneR = BASOP_Util_Add_Mant32Exp( eneR, eneR_e, W_round64_L( W_tmp ), sub( 1, L_tmp_e ), &eneR_e );
		#else
		Nr = BASOP_Util_Add_Mant32Exp( Nr, Nr_e, L_add( Mpy_32_32_r( specLr[i], tIPDr ), Mpy_32_32_r( specLi[i], tIPDi ) ), shl( spec_e, 1 ), &Nr_e );
		Ni = BASOP_Util_Add_Mant32Exp( Ni, Ni_e, L_sub( Mpy_32_32_r( specLi[i], tIPDr ), Mpy_32_32_r( specLr[i], tIPDi ) ), shl( spec_e, 1 ), &Ni_e );

		eneL = BASOP_Util_Add_Mant32Exp( eneL, eneL_e, L_add( Mpy_32_32_r( specLr[i], specLr[i] ), Mpy_32_32_r( specLi[i], specLi[i] ) ), shl( spec_e, 1 ), &eneL_e );
		eneR = BASOP_Util_Add_Mant32Exp( eneR, eneR_e, L_add( Mpy_32_32_r( specRr[i], specRr[i] ), Mpy_32_32_r( specRi[i], specRi[i] ) ), shl( spec_e, 1 ), &eneR_e );
		#endif
		}

		// Pn = (float) inv_sqrt( ( tPr * tPr + tPi * tPi ) + EPSILON );
		@@ -807,9 +865,19 @@ static void calc_poc_fx(
		move32();

		// Pn = (float) inv_sqrt( ( Pr[n] * Pr[n] + Pi[n] * Pi[n] ) + EPSILON );
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		Pn = L_add( L_shr( Mpy_32_32_r( Pr[n], Pr[n] ), 1 ), L_shr( Mpy_32_32_r( Pi[n], Pi[n] ), 1 ) );
		Pn = BASOP_Util_Add_Mant32Exp( Pn, 1, EPSILON_FX_M, EPSILON_FX_E, &Pn_e );
		Pn = Isqrt_lc( Pn, &Pn_e );
		#else
		L_tmp = L_add( L_shr( Mpy_32_32_r( Pr[n], Pr[n] ), 1 ), L_shr( Mpy_32_32_r( Pi[n], Pi[n] ), 1 ) );
		L_tmp_e = 1;
		move16();
		L_tmp = BASOP_Util_Add_Mant32Exp( L_tmp, L_tmp_e, EPSILON_FX_M, EPSILON_FX_E, &L_tmp_e );
		Pn_e = L_tmp_e;
		move16();
		Pn = ISqrt32( L_tmp, &Pn_e );
		#endif

		Pr[n] = L_shl_sat( Mpy_32_32_r( Pr[n], Pn ), Pn_e ); // Q31
		move32();
		@@ -1017,6 +1085,15 @@ static void calc_poc_fx(
		{
		// hPHA->p_curr_taps[n][i] *= hPHA->win[i];
		hPHA->p_curr_taps_fx[n][i] = Mpy_32_16_1( hPHA->p_curr_taps_fx[n][i], hPHA->win_fx[i] ); // Q30
		#else

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		hPHA->p_curr_taps_fx[n][i] = Mpy_32_16_1( hPHA->p_curr_taps_fx[n][i], hPHA->win_fx[i] ); // Q30
		#else
		hPHA->p_curr_taps_fx[n][i] = L_shl( Mpy_32_16_1( hPHA->p_curr_taps_fx[n][i], hPHA->win_fx[i] ), 1 ); // Q31
		#endif

		#endif
		move32();
		}

		@@ -1033,7 +1110,11 @@ static void calc_poc_fx(
		energy = ISqrt32( energy, &energy_e );
		FOR( i = 0; i < hPHA->pha_len; i++ )
		{
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		hPHA->p_curr_taps_fx[n][i] = L_shl_r( Mpy_32_32_r( hPHA->p_curr_taps_fx[n][i], energy ), sub( energy_e, 1 ) ); // Q30
		#else
		hPHA->p_curr_taps_fx[n][i] = L_shl_r( Mpy_32_32_r( hPHA->p_curr_taps_fx[n][i], energy ), energy_e ); // Q31
		#endif
		move32();
		}
		}
		@@ -1136,7 +1217,11 @@ static void calc_poc_fx(
		-------------------------------------------------------------------/
		static Word32 find_poc_peak_fx(
		STEREO_DMX_EVS_POC_HANDLE hPOC, /* i/o: phase only correlation structure */
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		Word16 itd_fx[], /* o : estimated itd Q0 */
		#else
		Word32 itd_fx[], /* o : estimated itd Q16 */
		#endif
		const Word16 input_frame, /* i : input frame length per channel */
		const Word32 ratio_fixed /* i : adapting ratio Q31 */
		)
		@@ -1448,7 +1533,11 @@ static ivas_error estimate_itd_fx(
		STEREO_DMX_EVS_PHA_HANDLE hPHA, /* i/o: correlation filter structure */
		const Word32 srcL[], /* i : Lch input signal Q16 */
		const Word32 srcR[], /* i : Rch input signal Q16 */
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		Word16 itd[], /* o : estimated itd Q0 */
		#else
		Word32 itd[], /* o : estimated itd Q16 */
		#endif
		const Word16 input_frame /* i : input frame length per channel */
		)
		{
		@@ -1989,7 +2078,9 @@ void stereo_dmx_evs_enc_fx(
		// ftmp += p_data_mem[n - m] * p_prev_taps[m];
		fx_tmp = L_add( fx_tmp, Mpy_32_32( p_data_mem[n - m], p_prev_taps[m] ) ); // Q25
		}
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		fx_tmp = L_shl( fx_tmp, 1 ); // Q26
		#endif
		mem_prev[n] = L_add( mem_prev[n], Mpy_32_32( fx_tmp, INV_SQRT_2_Q31 ) );
		move32();
		}
		@@ -2014,7 +2105,9 @@ void stereo_dmx_evs_enc_fx(
		// ftmp += p_data_mem[n - m] * p_curr_taps[m];
		fx_tmp = L_add( fx_tmp, Mpy_32_32( p_data_mem[n - m], p_curr_taps[m] ) ); // Q25
		}
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		fx_tmp = L_shl( fx_tmp, 1 ); // Q26
		#endif
		// dmx_pha_data[n] += ftmp * INV_SQRT_2;
		dmx_pha_data[n] = L_add( dmx_pha_data[n], Mpy_32_32( fx_tmp, INV_SQRT_2_Q31 ) ); // Q26
		move32();
		@@ -2044,7 +2137,11 @@ void stereo_dmx_evs_enc_fx(
		curr_prc = hStereoDmxEVS->hPHA->curr_prc;
		move32();
		// if ( abs( (int16_t) hStereoDmxEVS->itd ) > hStereoDmxEVS->hPHA->prc_thres )
		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		IF( GT_16( abs_s( hStereoDmxEVS->itd_fx ), hStereoDmxEVS->hPHA->prc_thres ) )
		#else
		IF( GT_16( abs_s( round_fx( hStereoDmxEVS->itd_fx ) ), hStereoDmxEVS->hPHA->prc_thres ) )
		#endif
		{
		IF( NE_32( hStereoDmxEVS->hPHA->curr_prc, STEREO_DMX_EVS_PRC_POC ) )
		{
		@@ -2401,6 +2498,7 @@ ivas_error stereo_dmx_evs_init_encoder_fx(
		fad_len = hStereoDmxEVS->hPHA->fad_len;
		move16();

		#ifdef FIX_1386_STEREO_DMX_EVS_PHA
		set16_fx( hStereoDmxEVS->hPHA->win_fx, 29491 /1.8f in Q14/, pha_len );
		hStereoDmxEVS->hPHA->win_fx[0] = ONE_IN_Q14;
		move16();
		@@ -2416,6 +2514,24 @@ ivas_error stereo_dmx_evs_init_encoder_fx(
		hStereoDmxEVS->hPHA->win_fx[sub( pha_len, 1 )] = 2816; /0.1718f in Q14/
		move16();
		}
		#else
		trans_len = idiv1616( pha_len, 20 );
		set16_fx( hStereoDmxEVS->hPHA->win_fx, 29491 /1.8f in Q15/, sub( pha_len, trans_len ) );
		hStereoDmxEVS->hPHA->win_fx[0] = ONE_IN_Q14;
		move16();
		IF( EQ_32( input_Fs, 16000 ) )
		{
		hStereoDmxEVS->hPHA->win_fx[pha_len - 1] = 7373; /0.45f in Q14/
		move16();
		}
		ELSE IF( EQ_32( input_Fs, 32000 ) \|\| EQ_32( input_Fs, 48000 ) )
		{
		hStereoDmxEVS->hPHA->win_fx[sub( pha_len, 2 )] = 19302; /1.1781f in Q14/
		move16();
		hStereoDmxEVS->hPHA->win_fx[sub( pha_len, 1 )] = 2816; /0.1718f in Q14/
		move16();
		}
		#endif

		fad_g = hStereoDmxEVS->hPHA->fad_g_fx;
		// fad_r = 1.0f / (float) ( fad_len + 1 );