Fix for 3GPP issue 1457: Stereo Encoder 32 kbps: More rattling artifcats in 16 kHz LTVs (d5d3ed62) · Commits · SA4 / Audio / IVAS BASOP

lib_enc/ivas_cpe_enc_fx.c

+6 −4

Original line number	Diff line number	Diff line
		@@ -649,10 +649,11 @@ ivas_error ivas_cpe_enc_fx(
		hCPE->hStereoDft->Spd_R_smooth_fx_e = sub( hCPE->hStereoDft->Spd_R_smooth_fx_e, temp );
		move16();


		q_com = MAX_16;
		move16();
		FOR( n = 0; n < CPE_CHANNELS; n++ )
		{
		q_com = sub( add( L_norm_arr( sts[n]->input32_fx, input_frame ), sts[n]->q_inp32 ), 16 );
		q_com = s_min( q_com, sub( add( L_norm_arr( sts[n]->input32_fx, input_frame ), sts[n]->q_inp32 ), 16 ) );
		q_com = s_min( q_com, sub( add( L_norm_arr( sts[n]->old_input_signal32_fx, input_frame ), sts[n]->q_old_inp32 ), 16 ) );
		q_com = s_min( q_com, add( norm_arr( hCPE->input_mem_fx[n], hCPE->hStereoDft->dft_ovl ), hCPE->q_input_mem[n] ) );

		@@ -661,7 +662,9 @@ ivas_error ivas_cpe_enc_fx(
		q_com = 0;
		move16();
		}

		}
		FOR( n = 0; n < CPE_CHANNELS; n++ )
		{
		Copy_Scale_sig_32_16( sts[n]->input32_fx, sts[n]->input_fx, input_frame, sub( q_com, sts[n]->q_inp32 ) );
		sts[n]->q_inp = q_com;
		move16();
		@@ -674,7 +677,6 @@ ivas_error ivas_cpe_enc_fx(
		hCPE->q_input_mem[n] = q_com;
		move16();
		}

		stereo_dft_hybrid_ITD_flag_fx( hCPE->hStereoDft->hConfig, input_Fs, hCPE->hStereoDft->hItd->hybrid_itd_max );

		/* Time Domain ITD compensation using extrapolation */

lib_enc/ivas_stereo_dft_enc_itd_fx.c

+24 −2

Original line number	Diff line number	Diff line
		@@ -169,7 +169,8 @@ static void stereo_dft_quantize_itd_fx(
		Word16 *ind )
		{
		Word16 itd;

		Word16 input_fs_by_stereo_dft_itd_fs = 0;
		move16();
		// itd = (int16_t) ( sign( in ) * 0.5f + in );
		itd = in;
		move16();
		@@ -188,9 +189,30 @@ static void stereo_dft_quantize_itd_fx(
		move16();
		}

		SWITCH( input_Fs )
		{
		case 8000:
		input_fs_by_stereo_dft_itd_fs = 1; // Q2
		move16();
		BREAK;
		case 16000:
		input_fs_by_stereo_dft_itd_fs = 2; // Q2
		move16();
		BREAK;
		case 32000:
		input_fs_by_stereo_dft_itd_fs = 4; // Q2
		move16();
		BREAK;
		case 48000:
		input_fs_by_stereo_dft_itd_fs = 6; // Q2
		move16();
		BREAK;
		default:
		assert( 0 );
		}
		/Convert back @ fs/
		//out = (float) ( itd input_Fs ) / ( (float) ( STEREO_DFT_ITD_FS ) );
		out = Mpy_32_32( L_shl( itd input_Fs, 2 ), 1099511628 /Q45/ ); // Q45+Q2-Q31 = Q16
		*out = L_shl( L_mult( itd, input_fs_by_stereo_dft_itd_fs ), Q13 ); // Q16
		move32();

		return;

lib_enc/ivas_stereo_dft_td_itd_fx.c

+49 −17

Original line number	Diff line number	Diff line
		@@ -50,7 +50,7 @@

		#define STEREO_DFT_CHANNEL_EXTR_LPC_ORDER 10
		#define STEREO_DFT_CHANNEL_EXTR_LPC_VEC_LIMIT 320
		#define FLR_FX 16384 // Q15
		#define FLR_FX 1073741824 // Q31
		#define ONE_HALF 24576 // Q14
		const Word16 wac_swb_h_fx[LPC_SHB_ORDER + 1] = { // Q15
		32767,
		@@ -95,7 +95,12 @@ static void stereo_td_get_td_itd_fx(
		{
		/* td_itd is the closest integer to itd that transforms into an integer value
		* under the transform x -> (32000/fs) x. */
		Word16 d, d_e;
		Word32 inverse_input_fs_by_16k = 0;
		move32();
		Word64 tmp64;
		Word16 d = 0, shift = 0;
		move16();
		move16();

		IF( EQ_32( input_Fs, 32000 ) )
		{
		@@ -106,25 +111,52 @@ static void stereo_td_get_td_itd_fx(
		ELSE
		{
		assert( ( input_Fs % 16000 ) == 0 && "sampling frequency should be divisible by 16000" );
		Word16 temp_div, temp_e, temp_add;
		d = BASOP_Util_Divide3232_Scale( input_Fs, 16000, &d_e );
		temp_div = BASOP_Util_Divide3232_Scale( itd, L_deposit_h( d ), &temp_e );
		temp_e = add( temp_e, sub( sub( 31, q_itd ), d_e ) ); // e+(e1-e2)//
		temp_add = add_sat( temp_div, shr_sat( FLR_FX, temp_e ) );

		IF( itd != 0 )
		IF( itd == 0 )
		{
		*td_itd_32 = extract_l( L_shl( L_shr( L_add( temp_add, EPSILON_FX ), sub( 15, temp_e ) ), 1 ) );
		td_itd_32 = td_itd = 0;
		move16();
		move16();
		}
		ELSE
		{
		*td_itd_32 = shl( shr( temp_add, sub( 15, temp_e ) ), 1 );
		SWITCH( input_Fs )
		{
		case 16000:
		inverse_input_fs_by_16k = 1073741824; // Q30 /* This is in Q30, because 2147483648 cannot be represnt in Word32*/
		move32();
		d = 1; // Q0
		move16();
		shift = 1;
		move16();
		BREAK;
		case 32000:
		inverse_input_fs_by_16k = 1073741824; // Q31
		move32();
		d = 2; // Q0
		move16();
		BREAK;
		case 48000:
		inverse_input_fs_by_16k = 715827883; // Q31
		move32();
		d = 3; // Q0
		move16();
		if (itd < 0 )
		{
		inverse_input_fs_by_16k = 715827882; /* Reducing the precision by 1 to handle multiplication of negative case*/
		move32();
		}
		BREAK;
		default:
		assert( 0 );
		}
		tmp64 = W_add( W_mult0_32_32( itd, inverse_input_fs_by_16k ), W_shl( FLR_FX, sub( q_itd, shift ) ) ); // Q: q_itd - shift + 31
		*td_itd_32 = shl( extract_l( W_extract_l( W_shr( tmp64, add( sub( q_itd, shift ), Q31 ) ) ) ), 1 ); // Q0
		move16();

		td_itd = i_mult( shr( td_itd_32, 1 ), shr( d, sub( 15, d_e ) ) );
		td_itd = i_mult( shr( td_itd_32, 1 ), d ); // Q0
		move16();
		}
		}

		return;
		}