Merge branch 'main' into update_crend_v2 (0db7946e) · Commits · SA4 / Audio / IVAS BASOP

apps/renderer.c

+50 −0

Original line number	Diff line number	Diff line
		@@ -662,6 +662,10 @@ int main(
		lfeRoutingConfigs[i] = NULL;
		}

		#ifdef FIX_DISCLAIMER
		IVAS_REND_PrintDisclaimer();

		#endif
		CmdlnArgs args = parseCmdlnArgs( argc, argv );

		if ( args.nonDiegeticPan && !( ( args.inConfig.numAudioObjects == 0 && args.inConfig.multiChannelBuses[0].audioConfig == IVAS_AUDIO_CONFIG_MONO ) \|\|
		@@ -829,6 +833,52 @@ int main(
		exit( -1 );
		}

		#ifdef FIX_DISCLAIMER
		fprintf( stdout, "Input audio file: %s\n", args.inputFilePath );
		fprintf( stdout, "Output audio file: %s\n\n", args.outputFilePath );

		if ( args.inConfig.numAudioObjects > 0 )
		{
		if ( args.inConfig.numAudioObjects == 1 )
		{
		IVAS_REND_PrintInputConfig( IVAS_AUDIO_CONFIG_ISM1 );
		}
		else if ( args.inConfig.numAudioObjects == 2 )
		{
		IVAS_REND_PrintInputConfig( IVAS_AUDIO_CONFIG_ISM2 );
		}
		else if ( args.inConfig.numAudioObjects == 3 )
		{
		IVAS_REND_PrintInputConfig( IVAS_AUDIO_CONFIG_ISM3 );
		}
		else if ( args.inConfig.numAudioObjects == 4 )
		{
		IVAS_REND_PrintInputConfig( IVAS_AUDIO_CONFIG_ISM4 );
		}
		}
		for ( i = 0; i < args.inConfig.numMultiChannelBuses; i++ )
		{
		IVAS_REND_PrintInputConfig( args.inConfig.multiChannelBuses[i].audioConfig );
		}
		for ( i = 0; i < args.inConfig.numMasaBuses; i++ )
		{
		IVAS_REND_PrintInputConfig( args.inConfig.masaBuses[i].audioConfig );
		}
		for ( i = 0; i < args.inConfig.numAmbisonicsBuses; i++ )
		{
		IVAS_REND_PrintInputConfig( args.inConfig.ambisonicsBuses[i].audioConfig );
		}

		if ( ( error = IVAS_REND_PrintConfig( hIvasRend ) ) != IVAS_ERR_OK )
		{
		fprintf( stderr, "\n IVAS_REND_PrintConfig failed: %s\n\n", ivas_error_to_string( error ) );
		// goto cleanup;
		exit( -1 );
		}

		#endif


		#ifdef FIX_POINT_HRTF_FILE_FORMAT
		if ( !isEmptyString( args.customHrtfFilePath ) )
		{

lib_com/disclaimer.c

+4 −0

Original line number	Diff line number	Diff line
		@@ -44,7 +44,11 @@ int16_t print_disclaimer( FILE *fPtr )
		{

		fprintf( fPtr, "\n==================================================================================================\n" );
		#ifdef FIX_DISCLAIMER
		fprintf( fPtr, " IVAS Codec BASOP Baseline\n" );
		#else
		fprintf( fPtr, " IVAS Codec Baseline\n" );
		#endif
		fprintf( fPtr, " \n" );
		fprintf( fPtr, " Based on EVS Codec (Floating Point) 3GPP TS26.443 Nov 04, 2021,\n" );
		fprintf( fPtr, " Version 12.14.0 / 13.10.0 / 14.6.0 / 15.4.0 / 16.3.0\n" );

lib_com/options.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -123,4 +123,5 @@
		#define FIX_ISSUE_1157 /* Ittiam: Fix for Issue 1157: Encoder crash for Stereo at 48/64kbps DTX on/off in kernel_switch_trafo_fx() */
		#define FIX_ISSUE_1152 /* Ittiam: Fix for issue 1152: Assertion error observed in evs_enc_fx (with option stereo_dmx_evs) from find_tilt_fx function*/
		#define FIX_ISSUE_1156 /* Ittiam: Fix for Issue 1156: Encoder crash for Stereo at 32kbps in SWB_BWE_encoding_ivas_fx() */
		#define FIX_DISCLAIMER /* VA: Add disclaimer for external renderer + Add info about IVAS reference version (FLP issue 1225) */
		#endif

lib_com/prot_fx.h

+17 −0

Original line number	Diff line number	Diff line
		@@ -2914,6 +2914,23 @@ void GenShapedWBExcitation_fx(
		const Word16 signal_type,
		const Word16 igf_flag );

		void GenShapedWBExcitation_ivas_enc_fx(
		Word16 excSHB, / o : synthesized shaped shb exctiation */
		const Word16 lpc_shb, / i : lpc coefficients */
		Word16 exc4kWhtnd, / o : whitened synthesized shb excitation */
		Word32 mem_csfilt, / i/o : memory */
		Word16 mem_genSHBexc_filt_down1, / i/o : memory */
		Word16 mem_genSHBexc_filt_down2, / i/o : memory */
		Word16 mem_genSHBexc_filt_down3, / i/o : memory */
		Word16 state_lpc_syn, / i/o : memory */
		const Word16 coder_type, /* i : coding type */
		const Word16 bwe_exc_extended, / i : bandwidth extended exciatation */
		const Word16 Q_bwe_exc,
		Word16 bwe_seed[], /* i/o : random number generator seed */
		const Word16 voice_factors[], /* i : voicing factor */
		const Word16 signal_type,
		const Word16 igf_flag );

		void GenWBSynth_fx(
		const Word16 input_synspeech, / i : i synthesized speech */
		Word16 shb_syn_speech_16k, / o : output highband compnent */

lib_com/swb_tbe_com_fx.c

+301 −0

Original line number	Diff line number	Diff line
		@@ -1951,6 +1951,307 @@ void GenShapedWBExcitation_fx(
		}


		void GenShapedWBExcitation_ivas_enc_fx(
		Word16 excSHB, / o : synthesized shaped shb exctiation Q(Q_bwe_exc) */
		const Word16 lpc_shb, / i : lpc coefficients Q12 */
		Word16 exc4kWhtnd, / o : whitened synthesized shb excitation Q(Q_bwe_exc) */
		Word32 mem_csfilt, / i/o : memory Q(Q_bwe_exc+16) */
		Word16 mem_genSHBexc_filt_down1, / i/o : memory Q(Q_bwe_exc) */
		Word16 mem_genSHBexc_filt_down2, / i/o : memory Q(Q_bwe_exc) */
		Word16 mem_genSHBexc_filt_down3, / i/o : memory Q(Q_bwe_exc) */
		Word16 state_lpc_syn, / i/o : memory Q(Q_bwe_exc) */
		const Word16 coder_type, /* i : coding type */
		const Word16 bwe_exc_extended, / i : bwidth extended exciatation Q(Q_bwe_exc) */
		const Word16 Q_bwe_exc, /* i : Q for memories */
		Word16 bwe_seed[], /* i/o : random number generator seed */
		const Word16 voice_factors[], /* i : voicing factor Q15 */
		const Word16 uv_flag, /* i : unvoiced flag */
		const Word16 igf_flag )
		{
		Word16 i, j, k;
		Word16 wht_fil_mem[LPC_WHTN_ORDER_WB];
		Word16 lpc_whtn[LPC_WHTN_ORDER_WB + 1];
		Word16 R_h[LPC_WHTN_ORDER_WB + 2], R_l[LPC_WHTN_ORDER_WB + 2];
		Word16 Q_R;
		Word16 excTmp[L_FRAME16k];
		Word16 excTmp2[L_FRAME16k / 4];
		Word16 excTmp2_frac[L_FRAME16k / 4];
		Word16 exc4k[L_FRAME16k / 4];
		Word16 exc4k_frac[L_FRAME16k / 4];
		Word32 exc4k_32[L_FRAME16k / 4];
		Word32 pow1, pow22;
		Word16 scale;
		Word32 excNoisyEnv[L_FRAME16k / 4];
		Word16 csfilt_num2[1] = { 1638 }; /* Q15*/
		Word16 neg_csfilt_den2[2] = { -32768, 31457 }; /* Q15 */
		Word32 L_tmp, Ltemp1, Ltemp2;
		Word16 temp1, temp2, exp;
		Word32 Lmax;
		Word16 max_val, n1, n2, sc;
		Word32 LepsP[LPC_WHTN_ORDER_WB + 1];
		Word16 tmp_vfac;
		Word16 avg_voice_fac;
		#ifdef BASOP_NOGLOB_DECLARE_LOCAL
		Flag Overflow = 0;
		move32();
		#endif
		move16();
		move16();
		move16();

		/0.25fsum_f(voice_factors, NB_SUBFR)*/
		L_tmp = L_mult( voice_factors[0], 8192 /0.25 in Q15 / );
		FOR( i = 1; i < NB_SUBFR; i++ )
		{
		L_tmp = L_mac( L_tmp, voice_factors[i], 8192 /0.25 in Q15 / );
		}
		avg_voice_fac = round_fx( L_tmp );

		test();
		test();
		test();
		test();
		IF( igf_flag != 0 && ( EQ_16( coder_type, VOICED ) \|\| GT_16( avg_voice_fac, 11469 /* 0.35 in Q15 / ) ) ) /Q15 -> 0.35f*/
		{
		csfilt_num2[0] = 6554;
		move16(); /Q15 -> 0.2f/
		neg_csfilt_den2[1] = 26214;
		move16(); /Q15 -> 0.8f/
		}
		ELSE IF( igf_flag != 0 && ( EQ_16( coder_type, UNVOICED ) \|\| LT_16( avg_voice_fac, 6654 /* 0.2 in Q15 / ) ) ) /Q15 -> 0.2f*/
		{
		csfilt_num2[0] = 328;
		move16(); /Q15 -> 0.01f/
		neg_csfilt_den2[1] = 32440;
		move16(); /Q15 -> 0.99f/
		}
		set16_fx( wht_fil_mem, 0, LPC_WHTN_ORDER_WB );
		Decimate_allpass_steep_fx( bwe_exc_extended, mem_genSHBexc_filt_down1, L_FRAME32k, excTmp );
		flip_spectrum_and_decimby4_fx( excTmp, exc4k, L_FRAME16k, mem_genSHBexc_filt_down2, mem_genSHBexc_filt_down3, 0 );

		IF( uv_flag )
		{
		create_random_vector_fx( exc4kWhtnd, L_FRAME16k / 4, bwe_seed );
		IF( LT_16( Q_bwe_exc, 5 ) )
		{
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		exc4kWhtnd[i] = shl_r( exc4kWhtnd[i], sub( Q_bwe_exc, 5 ) ); /Q(Q_bwe_exc)/Q5(if Q_bwe_exc > 5) /
		move16();
		}
		}
		}
		ELSE
		{
		autocorr_fx( exc4k, LPC_WHTN_ORDER_WB + 1, R_h, R_l, &Q_R,
		L_FRAME16k / 4, win_flatten_4k_fx, 0, 1 );

		/* Ensure R[0] isn't zero when entering Levinson Durbin */
		R_l[0] = s_max( R_l[0], 1 );
		move16();
		FOR( i = 1; i <= LPC_WHTN_ORDER_WB; i++ )
		{
		L_tmp = Mpy_32( R_h[i], R_l[i], wac_h[i - 1], wac_l[i - 1] );
		L_Extract( L_tmp, &R_h[i], &R_l[i] );
		}

		E_LPC_lev_dur( R_h, R_l, lpc_whtn, LepsP, LPC_WHTN_ORDER_WB, NULL );

		Copy_Scale_sig( lpc_whtn, lpc_whtn, LPC_WHTN_ORDER_WB + 1, sub( norm_s( lpc_whtn[0] ), 2 ) );

		fir_fx( exc4k, lpc_whtn, exc4kWhtnd, wht_fil_mem, L_FRAME16k / 4,
		LPC_WHTN_ORDER_WB, 0, 3 );

		/* Ensure pow1 is greater than zero when computing normalization */
		max_val = 0;
		move16();
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		excTmp2[i] = abs_s( exc4kWhtnd[i] );
		move16(); /* Q_bwe_exc */
		max_val = s_max( max_val, excTmp2[i] );
		move16();
		}

		IF( max_val == 0 )
		{
		pow1 = 1;
		move16();
		n1 = 0;
		move16();
		}
		ELSE
		{
		n1 = norm_s( max_val );
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		#ifdef BASOP_NOGLOB
		excTmp2_frac[i] = shl_o( excTmp2[i], n1, &Overflow );
		#else
		excTmp2_frac[i] = shl( excTmp2[i], n1 );
		#endif
		move16(); /* Q14 */
		}
		n1 = sub( sub( 14, n1 ), Q_bwe_exc );
		pow1 = 1;
		move32();
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		#ifdef BASOP_NOGLOB
		L_tmp = L_mult_o( excTmp2_frac[i], excTmp2_frac[i], &Overflow ); /* Q29 */
		pow1 = L_add_o( pow1, L_shr( L_tmp, 7 ), &Overflow ); /* Q22 */
		#else
		L_tmp = L_mult( excTmp2_frac[i], excTmp2_frac[i] ); /* Q29 */
		pow1 = L_add( pow1, L_shr( L_tmp, 7 ) ); /* Q22 */
		#endif
		}
		}

		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		#ifdef BASOP_NOGLOB
		excNoisyEnv[i] = L_add_o( *mem_csfilt, L_mult_o( csfilt_num2[0], excTmp2[i], &Overflow ), &Overflow );
		#else
		excNoisyEnv[i] = L_add( *mem_csfilt, L_mult( csfilt_num2[0], excTmp2[i] ) );
		#endif
		move32(); /* Q_bwe_exc+16 */
		*mem_csfilt = Mpy_32_16_1( excNoisyEnv[i], neg_csfilt_den2[1] );
		move32(); /* Q_bwe_exc+16 */
		}

		create_random_vector_fx( exc4k, L_FRAME16k / 4, bwe_seed );

		/* Ensure pow22 is greater than zero when computing normalization */
		Lmax = 0;
		move32();
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		exc4k_32[i] = Mpy_32_16_1( excNoisyEnv[i], exc4k[i] );
		move32(); /* Q_bwe_exc+6 */
		Lmax = L_max( Lmax, L_abs( exc4k_32[i] ) );
		}

		IF( Lmax == 0 )
		{
		pow22 = 1;
		move16();
		n2 = 0;
		move16();
		set16_fx( exc4k_frac, 0, L_FRAME16k / 4 );
		}
		ELSE
		{
		n2 = norm_l( Lmax );
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		#ifdef BASOP_NOGLOB
		exc4k_frac[i] = extract_h( L_shl_o( exc4k_32[i], n2, &Overflow ) ); /* Q(14-n2) */
		#else
		exc4k_frac[i] = extract_h( L_shl( exc4k_32[i], n2 ) ); /* Q(14-n2) */
		#endif
		move16();
		}
		n2 = sub( sub( 30, n2 ), add( Q_bwe_exc, 6 ) );
		pow22 = 1;
		move32();
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		#ifdef BASOP_NOGLOB
		L_tmp = L_mult_o( exc4k_frac[i], exc4k_frac[i], &Overflow ); /* Q29 */
		pow22 = L_add_o( pow22, L_shr( L_tmp, 7 ), &Overflow ); /* Q22 */
		#else
		L_tmp = L_mult( exc4k_frac[i], exc4k_frac[i] ); /* Q29 */
		pow22 = L_add( pow22, L_shr( L_tmp, 7 ) ); /* Q22 */
		#endif
		}
		}

		test();
		test();
		IF( EQ_16( coder_type, UNVOICED ) \|\| ( igf_flag != 0 && LT_16( avg_voice_fac, 6654 /0.2 in Q15 / ) ) )
		{
		L_tmp = root_a_over_b_fx( pow1, sub( 22, shl( n1, 1 ) ), pow22, sub( 22, shl( n2, 1 ) ), &exp );
		#ifdef BASOP_NOGLOB
		scale = round_fx_o( L_shl_o( L_tmp, exp, &Overflow ), &Overflow ); /Q15 /
		#else
		scale = round_fx( L_shl( L_tmp, exp ) ); /Q15 /
		#endif
		sc = sub( add( n2, Q_bwe_exc ), 14 );
		FOR( i = 0; i < L_FRAME16k / 4; i++ )
		{
		#ifdef BASOP_NOGLOB
		exc4kWhtnd[i] = round_fx_o( L_shl_o( L_mult_o( exc4k_frac[i], scale, &Overflow ), sc, &Overflow ), &Overflow ); /* Q_bwe_exc+n2-10+16+ Q_bwe_exc + n2 -14 -16 = //Q_bwe_exc */
		#else
		exc4kWhtnd[i] = round_fx( L_shl( L_mult( exc4k_frac[i], scale ), sc ) ); /* Q_bwe_exc+n2-10+16+ Q_bwe_exc + n2 -14 -16 = //Q_bwe_exc */
		#endif
		move16();
		}
		}
		ELSE
		{
		sc = sub( add( n2, Q_bwe_exc ), 14 ); /* Q_bwe_exc+n2-14*/

		k = 0;
		move16();
		FOR( i = 0; i < 4; i++ )
		{
		test();
		IF( igf_flag != 0 && EQ_16( coder_type, VOICED ) )
		{
		/tmp_vfac = 2voice_factors[i];
		tmp_vfac = min(1, tmp_vfac);*/
		#ifdef BASOP_NOGLOB
		tmp_vfac = shl_o( voice_factors[i], 1, &Overflow );
		#else
		BASOP_SATURATE_WARNING_OFF_EVS
		tmp_vfac = shl( voice_factors[i], 1 );
		BASOP_SATURATE_WARNING_ON_EVS
		#endif
		}
		ELSE
		{
		tmp_vfac = voice_factors[i];
		move16();
		}

		Ltemp1 = root_a_fx( L_deposit_h( tmp_vfac ), 31, &exp );
		#ifdef BASOP_NOGLOB
		temp1 = round_fx_o( L_shl_o( Ltemp1, exp, &Overflow ), &Overflow ); /* Q15 */
		#else
		temp1 = round_fx( L_shl( Ltemp1, exp ) ); /* Q15 */
		#endif
		L_tmp = Mpy_32_16_1( pow1, sub( 32767, tmp_vfac ) ); /* Q22*/
		Ltemp2 = root_a_over_b_fx( L_tmp, sub( 22, shl( n1, 1 ) ), pow22, sub( 22, shl( n2, 1 ) ), &exp );
		#ifdef BASOP_NOGLOB
		temp2 = round_fx_o( L_shl_o( Ltemp2, exp, &Overflow ), &Overflow ); /* Q15 */
		#else
		temp2 = round_fx( L_shl( Ltemp2, exp ) ); /* Q15 */
		#endif
		FOR( j = 0; j < L_FRAME16k / 16; j++ )
		{
		#ifdef BASOP_NOGLOB
		L_tmp = L_mult_o( temp1, exc4kWhtnd[k], &Overflow ); /* Q(16+Q_bwe_exc) */
		L_tmp = L_add_o( L_tmp, L_shl_o( L_mult_o( temp2, exc4k_frac[k], &Overflow ), sc, &Overflow ), &Overflow ); /* Q(16+Q_bwe_exc) */
		exc4kWhtnd[k] = round_fx_o( L_tmp, &Overflow ); /* Q_bwe_exc */
		#else
		L_tmp = L_mult( temp1, exc4kWhtnd[k] ); /* Q(16+Q_bwe_exc) */
		L_tmp = L_add( L_tmp, L_shl( L_mult( temp2, exc4k_frac[k] ), sc ) ); /* Q(16+Q_bwe_exc) */
		exc4kWhtnd[k] = round_fx( L_tmp ); /* Q_bwe_exc */
		#endif
		move16();
		k = add( k, 1 );
		}
		}
		}
		}

		syn_filt_fx( 0, lpc_shb, LPC_SHB_ORDER_WB, exc4kWhtnd, excSHB, L_FRAME16k / 4, state_lpc_syn, 1 );


		return;
		}


		/-------------------------------------------------------------------
		* GenWBSynth()
		*