harmonize wb_bwe_enc_fx() (3b48390e) · Commits · SA4 / Audio / IVAS BASOP

lib_com/options.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -86,6 +86,7 @@
		#define FIX_1904_HARM_GSC_ENC /* VA: #1904 Harmonization of EVS and IVAS GSC code */
		#define FIX_2397_COPY_AQ_MDCT_CORE_BFI /* FhG: prevent copying of uninit memory in MDCT stereo core if bfi is set */
		#define HARMONIZE_TBE /* VA: harmonize core-coder TBE function duplications */
		#define HARM_FD_BWE /* VA: harmonize core-coder FD BWE function duplications */

		/* #################### End BE switches ################################## */

lib_enc/evs_enc_fx.c

+4 −0

Original line number	Diff line number	Diff line
		@@ -438,7 +438,11 @@ ivas_error evs_enc_fx(
		ELSE IF( EQ_16( st->extl, WB_BWE ) )
		{
		/* WB BWE encoder */
		#ifdef HARM_FD_BWE
		wb_bwe_enc_fx( st, new_inp_resamp16k );
		#else
		wb_bwe_enc_fx( st, new_inp_resamp16k, st->coder_type );
		#endif
		}

		/---------------------------------------------------------------------

lib_enc/ivas_core_enc_fx.c

+4 −0

Original line number	Diff line number	Diff line
		@@ -645,7 +645,11 @@ ivas_error ivas_core_enc_fx(
		ELSE IF( EQ_16( st->extl, WB_BWE ) && n == 0 && NE_16( st->element_mode, IVAS_CPE_MDCT ) )
		{
		/* WB BWE encoder */
		#ifdef HARM_FD_BWE
		wb_bwe_enc_fx( st, new_inp_resamp16k_fx[n] );
		#else
		wb_bwe_enc_ivas_fx( st, new_inp_resamp16k_fx[n] );
		#endif
		}

		/---------------------------------------------------------------------

lib_enc/prot_fx_enc.h

+2 −17

Original line number	Diff line number	Diff line
		@@ -669,12 +669,14 @@ void vad_param_updt_fx(
		);

		void wb_bwe_enc_fx(
		#ifndef HARM_FD_BWE
		Encoder_State st_fx, / i/o: encoder state structure */
		const Word16 new_wb_speech_fx, / i : original i signal at 16kHz */
		Word16 coder_type /* i : coding type */
		);

		void wb_bwe_enc_ivas_fx(
		#endif
		Encoder_State st_fx, / i/o: encoder state structure */
		const Word16 new_wb_speech_fx / i : original i signal at 16kHz */
		);
		@@ -981,23 +983,6 @@ void set_bw_stereo_fx(
		CPE_ENC_HANDLE hCPE /* i/o: CPE encoder structures */
		);

		/* o : classification of wb signal */
		Word16 WB_BWE_encoding_fx(
		const Word16 coder_type, /* i : coder type */
		const Word16 yos_fx, / i : MDCT coefficients of weighted original */
		Word16 WB_fenv_fx, / i/o: energy of WB envelope */
		Encoder_State st_fx, / i/o: Encoder structure */
		Word16 Q_synth,
		Word16 Q_synth_lf );

		/* o : classification of wb signal */
		Word16 WB_BWE_encoding_ivas_fx(
		Encoder_State st_fx, / i/o: Encoder structure */
		const Word32 yos_fx, / i : MDCT coefficients of weighted original */
		Word16 WB_fenv_fx, / i/o: energy of WB envelope */
		Word16 Q_synth,
		Word16 Q_synth_lf );

		void calculate_hangover_attenuation_gain_fx(
		Encoder_State st, / i : encoder state structure */
		Word16 att, / o : attenuation factor */

lib_enc/swb_bwe_enc_fx.c

+114 −73

Original line number	Diff line number	Diff line
		@@ -11,39 +11,23 @@
		#include "prot_fx_enc.h" /* Function prototypes */

		/---------------------------------------------------------------------
		* Local functions
		* Local functions declarations
		---------------------------------------------------------------------/

		static Word16 SWB_BWE_encoding_fx(
		Encoder_State st_fx, / i/o: encoder state structure */
		Word16 insig_fx, / i/o: delayed original input signal at 32kHz (might be rescaled)*/
		const Word16 insig_lp_fx, / i : delayed original lowband input signal at 32kHz */
		const Word16 insig_hp_fx, / i : delayed original highband input signal at 32kHz */
		const Word16 synth_fx, / i : delayed ACELP core synthesis at 12.8kHz */
		const Word16 yos_fx, / i : MDCT coefficients of the windowed original input signal at 32kHz */
		Word16 SWB_fenv_fx, / o : frequency-domain quantized BWE envelope */
		const Word16 tilt_nb_fx, /* i : SWB tilt */
		const Word16 st_offset, /* i : start frequency offset for BWE envelope */
		const Word16 coder_type, /* i : coding type */
		Word16 Q_insig_lp,
		Word16 Q_shb,
		Word16 Q_synth,
		Word16 Q_synth_lf );
		static Word16 WB_BWE_encoding_fx( const Word16 coder_type, const Word16 yos_fx, Word16 WB_fenv_fx, Encoder_State *st_fx, const Word16 Q_synth, const Word16 Q_synth_lf );

		static Word16 SWB_BWE_encoding_ivas_fx(
		Encoder_State st_fx, / i/o: encoder state structure */
		Word16 insig_fx, / i/o: delayed original input signal at 32kHz (might be rescaled)*/
		const Word16 insig_lp_fx, / i : delayed original lowband input signal at 32kHz */
		const Word16 insig_hp_fx, / i : delayed original highband input signal at 32kHz */
		const Word16 synth_fx, / i : delayed ACELP core synthesis at 12.8kHz */
		const Word32 yos_fx, / i : MDCT coefficients of the windowed original input signal at 32kHz */
		Word16 SWB_fenv_fx, / o : frequency-domain quantized BWE envelope */
		const Word16 tilt_nb_fx, /* i : SWB tilt */
		const Word16 st_offset, /* i : start frequency offset for BWE envelope */
		Word16 Q_insig_lp,
		Word16 Q_shb,
		Word16 Q_synth,
		Word16 Q_synth_lf );
		static Word16 WB_BWE_encoding_fx_32( Encoder_State st_fx, const Word32 yos_fx, Word16 *WB_fenv_fx, const Word16 Q_synth, const Word16 Q_synth_lf );

		static Word16 SWB_BWE_encoding_fx( Encoder_State st_fx, Word16 insig_fx, const Word16 insig_lp_fx, const Word16 insig_hp_fx, const Word16 synth_fx, const Word16 yos_fx, Word16 *SWB_fenv_fx, const Word16 tilt_nb_fx, const Word16 st_offset, const Word16 coder_type, const Word16 Q_insig_lp, const Word16 Q_shb, const Word16 Q_synth, const Word16 Q_synth_lf );

		static Word16 SWB_BWE_encoding_fx_32( Encoder_State st_fx, Word16 insig_fx, const Word16 insig_lp_fx, const Word16 insig_hp_fx, const Word16 synth_fx, const Word32 yos_fx, Word16 *SWB_fenv_fx, const Word16 tilt_nb_fx, const Word16 st_offset, const Word16 Q_insig_lp, const Word16 Q_shb, const Word16 Q_synth, const Word16 Q_synth_lf );


		/-------------------------------------------------------------------
		* delay_input_signal_fx()
		*
		*
		-------------------------------------------------------------------/

		static void delay_input_signal_fx(
		Word16 *old_sig,
		@@ -115,6 +99,7 @@ static void delay_input_signal_fx(
		return;
		}


		/-------------------------------------------------------------------
		* wb_bwe_enc()
		*
		@@ -122,6 +107,7 @@ static void delay_input_signal_fx(
		-------------------------------------------------------------------/

		void wb_bwe_enc_fx(
		#ifndef HARM_FD_BWE
		Encoder_State st_fx, / i/o: encoder state structure */
		const Word16 new_wb_speech_fx, / i : original input signal at 16kHz */
		Word16 coder_type /* i : coding type */
		@@ -201,6 +187,7 @@ void wb_bwe_enc_fx(
		-------------------------------------------------------------------/

		void wb_bwe_enc_ivas_fx(
		#endif
		Encoder_State st_fx, / i/o: encoder state structure */
		const Word16 new_wb_speech_fx / i : original input signal at 16kHz */
		)
		@@ -232,12 +219,18 @@ void wb_bwe_enc_ivas_fx(

		/---------------------------------------------------------------------/
		/* WB BWE encoding */


		/* MDCT of the core synthesis signal */
		/---------------------------------------------------------------------/

		new_input_fx_exp = -1;
		move16();
		#ifdef HARM_FD_BWE
		if ( st_fx->element_mode == EVS_MONO )
		{
		new_input_fx_exp = 0;
		move16();
		}
		#endif

		wtda_fx( old_input_fx, &new_input_fx_exp, L_wtda_synth_fx, hBWE_FD->L_old_wtda_swb_fx,
		&st_fx->Q_old_wtda, ALDO_WINDOW, ALDO_WINDOW, /* window overlap of current frame (0: full, 2: none, or 3: half) */
		@@ -246,8 +239,37 @@ void wb_bwe_enc_ivas_fx(
		/* DCT of the ACELP core synthesis */
		direct_transform_fx( L_wtda_synth_fx, yorig_32, 0, L_FRAME16k, &new_input_fx_exp, st_fx->element_mode );

		mode = WB_BWE_encoding_ivas_fx( st_fx, yorig_32, WB_fenv_fx, new_input_fx_exp, new_input_fx_exp );
		#ifdef HARM_FD_BWE
		IF( st_fx->element_mode == EVS_MONO )
		{
		Word16 scl, Q_synth;
		Word16 yorig_fx[L_FRAME16k];

		/* Convert to 16 Bits (Calc Shift Required to Stay within MAX_Q_NEW_INPUT) */
		scl = sub( 16 + 8 /MAX_Q_NEW_INPUT/, new_input_fx_exp );

		/* Possible to Upscale? */
		IF( scl > 0 )
		{
		/* Yes */
		/* Calc Room to Upscale */
		Q_synth = Find_Max_Norm32( yorig_32, L_FRAME16k );

		/* Stay within MAX_Q_NEW_INPUT */
		scl = s_min( Q_synth, scl );
		}
		Copy_Scale_sig32_16( yorig_32, yorig_fx, L_FRAME16k, scl );
		Q_synth = sub( add( sub( new_input_fx_exp, 16 ), scl ), 1 );

		mode = WB_BWE_encoding_fx( st_fx->coder_type, yorig_fx, WB_fenv_fx, st_fx, Q_synth, Q_synth );
		}
		ELSE
		#endif
		{
		mode = WB_BWE_encoding_fx_32( st_fx, yorig_32, WB_fenv_fx, new_input_fx_exp, new_input_fx_exp );
		}
		move16();

		push_indice( st_fx->hBstr, IND_WB_CLASS, sub( mode, 2 ), 1 );
		}

		@@ -559,12 +581,12 @@ void swb_bwe_enc_ivas_fx(
		/* SWB BWE encoding */
		IF( EQ_16( st_fx->L_frame, L_FRAME16k ) )
		{
		SWB_BWE_encoding_ivas_fx( st_fx, old_input_fx, old_input_lp_fx, new_input_hp_fx, old_syn_12k8_16k_fx, yorig_32,
		SWB_BWE_encoding_fx_32( st_fx, old_input_fx, old_input_lp_fx, new_input_hp_fx, old_syn_12k8_16k_fx, yorig_32,
		SWB_fenv_fx, tilt_nb_fx, 80, Q_slb_speech, Q_shb, new_input_fx_q, new_input_fx_q );
		}
		ELSE
		{
		SWB_BWE_encoding_ivas_fx( st_fx, old_input_fx, old_input_lp_fx, new_input_hp_fx, old_syn_12k8_16k_fx, yorig_32,
		SWB_BWE_encoding_fx_32( st_fx, old_input_fx, old_input_lp_fx, new_input_hp_fx, old_syn_12k8_16k_fx, yorig_32,
		SWB_fenv_fx, tilt_nb_fx, 6, Q_slb_speech, Q_shb, new_input_fx_q, new_input_fx_q );
		}

		@@ -2048,6 +2070,7 @@ static void calculate_Tonality_fx(
		return;
		}

		#ifndef HARM_FD_BWE
		/-------------------------------------------------------------------
		* calculate_Tonality_ivas_fx()
		*
		@@ -2201,6 +2224,7 @@ static void calculate_Tonality_ivas_fx(

		return;
		}
		#endif

		/-------------------------------------------------------------------
		* energy_control_fx()
		@@ -2294,6 +2318,7 @@ static void energy_control_fx(
		return;
		}

		#ifndef HARM_FD_BWE
		/-------------------------------------------------------------------
		* energy_control_ivas_fx()
		*
		@@ -2386,6 +2411,7 @@ static void energy_control_ivas_fx(
		}
		return;
		}
		#endif

		/-------------------------------------------------------------------
		* WB_BWE_encoding()
		@@ -2393,13 +2419,14 @@ static void energy_control_ivas_fx(
		* WB BWE main encoder
		-------------------------------------------------------------------/

		Word16 WB_BWE_encoding_fx( /* o : classification of wb signal */
		/* o : classification of wb signal */
		static Word16 WB_BWE_encoding_fx(
		const Word16 coder_type, /* i : coder type */
		const Word16 yos_fx, / i : MDCT coefficients of weighted original */
		Word16 WB_fenv_fx, / i/o: energy of WB envelope */
		Encoder_State st_fx, / i/o: Encoder structure */
		Word16 Q_synth,
		Word16 Q_synth_lf )
		const Word16 Q_synth,
		const Word16 Q_synth_lf )
		{
		Word16 mode;
		Word16 i, n_coeff, n_band;
		@@ -2459,17 +2486,18 @@ Word16 WB_BWE_encoding_fx( /* o : classification of wb
		}

		/-------------------------------------------------------------------
		* WB_BWE_encoding_ivas()
		* WB_BWE_encoding()
		*
		* WB BWE main encoder
		* WB BWE main encoder, 32-bit variant
		-------------------------------------------------------------------/

		Word16 WB_BWE_encoding_ivas_fx( /* o : classification of wb signal */
		/* o : classification of wb signal */
		static Word16 WB_BWE_encoding_fx_32(
		Encoder_State st_fx, / i/o: Encoder structure */
		const Word32 yos_fx, / i : MDCT coefficients of weighted original */
		Word16 WB_fenv_fx, / i/o: energy of WB envelope */
		Word16 Q_synth,
		Word16 Q_synth_lf )
		const Word16 Q_synth,
		const Word16 Q_synth_lf )
		{
		Word16 mode;
		Word16 i, n_coeff, n_band;
		@@ -2484,6 +2512,7 @@ Word16 WB_BWE_encoding_ivas_fx( /* o : classification of
		Word16 q_shift, scale;
		Word16 q_WB_fenv[2];
		Word16 yos_fx_16[L_FRAME16k];

		n_band = 0;
		move16();
		FOR( i = 0; i < 2; i++ )
		@@ -2510,7 +2539,11 @@ Word16 WB_BWE_encoding_ivas_fx( /* o : classification of

		mode = FD_BWE_class_fx( yos_fx_16, 0, 0, sub( add( Q_synth, scale ), Q16 ), 0, st_fx );

		#ifdef HARM_FD_BWE
		energy_control_fx( st_fx, ACELP_CORE, mode, st_fx->coder_type, yos_fx_16, 0, energy_factor_fx, sub( add( Q_synth_lf, scale ), Q16 ) );
		#else
		energy_control_ivas_fx( st_fx, ACELP_CORE, mode, st_fx->coder_type, yos_fx_16, 0, energy_factor_fx, sub( add( Q_synth_lf, scale ), Q16 ) );
		#endif

		FOR( i = 0; i < 2; i++ )
		{
		@@ -2545,6 +2578,7 @@ Word16 WB_BWE_encoding_ivas_fx( /* o : classification of
		*
		* SWB BWE encoder
		-------------------------------------------------------------------/

		static Word16 SWB_BWE_encoding_fx(
		Encoder_State st_fx, / i/o: encoder state structure */
		Word16 insig_fx, / i : delayed original input signal at 32kHz */
		@@ -2556,10 +2590,10 @@ static Word16 SWB_BWE_encoding_fx(
		const Word16 tilt_nb_fx, /* i : SWB tilt */
		const Word16 st_offset, /* i : start frequency offset for BWE envelope */
		const Word16 coder_type, /* i : coding type */
		Word16 Q_insig_lp,
		Word16 Q_shb,
		Word16 Q_synth,
		Word16 Q_synth_lf )
		const Word16 Q_insig_lp,
		const Word16 Q_shb,
		const Word16 Q_synth,
		const Word16 Q_synth_lf )
		{
		Word16 IsTransient, mode;
		Word16 index;
		@@ -2982,9 +3016,10 @@ static Word16 SWB_BWE_encoding_fx(
		/-------------------------------------------------------------------
		* SWB_BWE_encoding()
		*
		* SWB BWE encoder
		* SWB BWE encoder, 32-bit variant
		-------------------------------------------------------------------/
		static Word16 SWB_BWE_encoding_ivas_fx(

		static Word16 SWB_BWE_encoding_fx_32(
		Encoder_State st_fx, / i/o: encoder state structure */
		Word16 insig_fx, / i : delayed original input signal at 32kHz */
		const Word16 insig_lp_fx, / i : delayed original lowband input signal at 32kHz */
		@@ -2994,10 +3029,10 @@ static Word16 SWB_BWE_encoding_ivas_fx(
		Word16 SWB_fenv_fx, / o : frequency-domain quantized BWE envelope */
		const Word16 tilt_nb_fx, /* i : SWB tilt */
		const Word16 st_offset, /* i : start frequency offset for BWE envelope */
		Word16 Q_insig_lp,
		Word16 Q_shb,
		Word16 Q_synth,
		Word16 Q_synth_lf )
		const Word16 Q_insig_lp,
		const Word16 Q_shb,
		const Word16 Q_synth,
		const Word16 Q_synth_lf )
		{
		Word16 IsTransient, mode;
		Word16 index;
		@@ -3411,7 +3446,11 @@ static Word16 SWB_BWE_encoding_ivas_fx(
		mode = FD_BWE_class_fx( yos_fx_16, global_gain_fx, tilt_nb_fx, sub( add( Q_synth, scale ), Q16 ), Q_shb, st_fx );
		push_indice( hBstr, IND_SWB_CLASS, mode, 2 );

		#ifdef HARM_FD_BWE
		energy_control_fx( st_fx, ACELP_CORE, mode, -1, yos_fx_16, st_offset, energy_factor_fx, sub( add( Q_synth_lf, scale ), Q16 ) );
		#else
		energy_control_ivas_fx( st_fx, ACELP_CORE, mode, -1, yos_fx_16, st_offset, energy_factor_fx, sub( add( Q_synth_lf, scale ), Q16 ) );
		#endif

		FOR( n_band = 0; n_band < SWB_FENV; n_band++ )
		{
		@@ -3555,6 +3594,7 @@ static void get_normalize_spec_fx_32(
		* calculate_tonality_fx_32()
		*
		-------------------------------------------------------------------/

		static void calculate_tonality_fx_32(
		const Word32 org_fx, / i : MDCT coefficients of original : Q12 */
		const Word32 gen_fx, / i : MDCT coefficients of generated signal : Q12 */
		@@ -3687,6 +3727,7 @@ static void calculate_tonality_fx_32(
		* energy_control_fx_32()
		*
		-------------------------------------------------------------------/

		static void energy_control_fx_32(
		Encoder_State st_fx, / i/o: encoder structure */
		const Word16 core, /* i : core : Q0 */
		@@ -3870,6 +3911,7 @@ static Word16 decision_hq_generic_class_fx_32(
		* hq_generic_encoding_fx()
		*
		-------------------------------------------------------------------/

		void hq_generic_encoding_fx(
		const Word32 coefs_fx, / i : MDCT coefficients of weighted original : Q12 */
		Word16 hq_generic_fenv_fx, / i/o: energy of SWB envelope : Q3 */
		@@ -3903,7 +3945,6 @@ void hq_generic_encoding_fx(
		move16();
		}


		energy_control_fx_32( st_fx, HQ_CORE, -1, -1, coefs_fx, hq_generic_offset, energy_factor_fx );

		IF( EQ_16( hHQ_core->hq_generic_speech_class, 1 ) )
		@@ -4021,7 +4062,6 @@ void hq_generic_encoding_fx(
		}
		}


		/* Energy VQ */
		IF( LE_16( hq_generic_offset, HQ_GENERIC_FOFFSET_24K4 ) )
		{
		@@ -4088,6 +4128,7 @@ void hq_generic_encoding_fx(
		return;
		}


		/-------------------------------------------------------------------
		* fd_bwe_enc_init_fx()
		*
		@@ -4134,10 +4175,12 @@ void fd_bwe_enc_init_fx(
		return;
		}


		/-------------------------------------------------------------------
		* hq_generic_encoding_fx()
		* hq_generic_hf_encoding_fx()
		*
		-------------------------------------------------------------------/

		void hq_generic_hf_encoding_fx(
		const Word32 coefs_fx, / i : MDCT coefficients of weighted original : Q12 */
		Word16 hq_generic_fenv_fx, / i/o: energy of SWB envelope : Q1 */
		@@ -4172,7 +4215,6 @@ void hq_generic_hf_encoding_fx(
		move16();
		}


		energy_control_fx_32( st_fx, HQ_CORE, -1, -1, coefs_fx, hq_generic_offset, energy_factor_fx );

		IF( EQ_16( hHQ_core->hq_generic_speech_class, 1 ) )
		@@ -4290,7 +4332,6 @@ void hq_generic_hf_encoding_fx(
		}
		}


		/* Energy VQ */
		IF( LE_16( hq_generic_offset, HQ_GENERIC_FOFFSET_24K4 ) )
		{