harmonize encod_gen_voic_fx(), under CLEANUP_ACELP_ENC (08a75c40) · Commits · SA4 / Audio / IVAS BASOP

lib_enc/acelp_core_enc_fx.c

+18 −0

Original line number	Diff line number	Diff line
		@@ -528,8 +528,13 @@ ivas_error acelp_core_enc_fx(
		hTdCngEnc->burst_ho_cnt = 0;
		move16();
		/* VOICED frames in SC-VBR when bumped up*/
		#ifdef CLEANUP_ACELP_ENC
		encod_gen_voic_fx( st_fx, inp_fx, Aw_fx, Aq_fx, Es_pred_fx, res_fx, syn_fx,
		exc_fx, exc2_fx, pitch_buf_fx, voice_factors_fx, bwe_exc_fx, unbits_fx, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf, shift, Q_new );
		#else
		encod_gen_voic_fx( st_fx, inp_fx, Aw_fx, Aq_fx, Es_pred_fx, res_fx, syn_fx,
		exc_fx, exc2_fx, pitch_buf_fx, voice_factors_fx, bwe_exc_fx, unbits_fx, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf, shift, Q_new );
		#endif
		}
		}
		ELSE IF( EQ_16( coder_type, AUDIO ) \|\| ( EQ_16( coder_type, INACTIVE ) && LE_32( st_fx->core_brate, ACELP_24k40 ) ) )
		@@ -541,8 +546,13 @@ ivas_error acelp_core_enc_fx(
		ELSE
		{
		/* GENERIC, VOICED and INACTIVE frames (coded by AVQ technology) */
		#ifdef CLEANUP_ACELP_ENC
		encod_gen_voic_fx( st_fx, inp_fx, Aw_fx, Aq_fx, Es_pred_fx, res_fx, syn_fx,
		exc_fx, exc2_fx, pitch_buf_fx, voice_factors_fx, bwe_exc_fx, unbits_fx, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf, shift, Q_new );
		#else
		encod_gen_voic_fx( st_fx, inp_fx, Aw_fx, Aq_fx, Es_pred_fx, res_fx, syn_fx,
		exc_fx, exc2_fx, pitch_buf_fx, voice_factors_fx, bwe_exc_fx, unbits_fx, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf, shift, Q_new );
		#endif
		}

		/* update st->mem_syn1 for ACELP core switching */
		@@ -1385,7 +1395,11 @@ ivas_error acelp_core_enc_ivas_fx(
		move16();

		/* VOICED frames in SC-VBR */
		#ifdef CLEANUP_ACELP_ENC
		encod_gen_voic_fx( st, inp, Aw, Aq, Es_pred_fx, res_fx, syn_fx, exc_fx, exc2_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, unbits, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf_fx, 0, Q_new );
		#else
		encod_gen_voic_ivas_fx( st, inp, Aw, Aq, Es_pred_fx, res_fx, syn_fx, exc_fx, exc2_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, unbits, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf_fx, 0, Q_new );
		#endif
		}
		}
		ELSE IF( EQ_16( st->coder_type, AUDIO ) \|\| ( st->coder_type == INACTIVE && st->inactive_coder_type_flag ) )
		@@ -1400,7 +1414,11 @@ ivas_error acelp_core_enc_ivas_fx(
		ELSE
		{
		/* GENERIC, VOICED and INACTIVE frames (coded by AVQ technology) */
		#ifdef CLEANUP_ACELP_ENC
		encod_gen_voic_fx( st, inp, Aw, Aq, Es_pred_fx, res_fx, syn_fx, exc_fx, exc2_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, unbits, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf_fx, 0, Q_new );
		#else
		encod_gen_voic_ivas_fx( st, inp, Aw, Aq, Es_pred_fx, res_fx, syn_fx, exc_fx, exc2_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, unbits, tdm_Pitch_reuse_flag, tdm_Pri_pitch_buf_fx, 0, Q_new );
		#endif
		}

		/* update mem_syn1_flt for ACELP core switching */

lib_enc/enc_gen_voic_fx.c

+119 −53

Original line number	Diff line number	Diff line
		@@ -77,7 +77,7 @@ static void rem_offset(
		/* RETURN ARGUMENTS : */
		/* _ None */
		/======================================================================/

		#ifndef CLEANUP_ACELP_ENC
		void encod_gen_voic_fx(
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx[], /* i : input speech Q0*/
		@@ -444,18 +444,22 @@ void encod_gen_voic_fx(

		return;
		}
		#endif


		#ifdef CLEANUP_ACELP_ENC
		void encod_gen_voic_fx(
		#else
		void encod_gen_voic_ivas_fx(
		#endif
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx[], /* i : input speech Qnew -1 */
		const Word16 speech_fx[], /* i : input speech Q0 / Qnew -1 */
		const Word16 Aw_fx[], /* i : weighted A(z) unquantized for subframes Q12*/
		const Word16 Aq_fx[], /* i : 12k8 Lp coefficient Q12*/
		const Word16 Es_pred_fx, /* i : predicted scaled innov. energy Q8*/
		const Word16 res_fx, / i : residual signal Q_new*/
		Word16 syn_fx, / i/o: core synthesis Q_new - 1*/
		Word16 exc_fx, / i/o: current non-enhanced excitation Q_new*/
		Word16 exc2_fx, / i/o: current enhanced excitation Q_new*/
		Word16 syn_fx, / i/o: core synthesis Qnew / Q_new - 1*/
		Word16 exc_fx, / i/o: current non-enhanced excitation Q0 / Q_new*/
		Word16 exc2_fx, / i/o: current enhanced excitation Q0 / Q_new*/
		Word16 pitch_buf_fx, / i/o: floating pitch values for each subframe Q6*/
		Word16 voice_factors_fx, / o : voicing factors Q15*/
		Word16 bwe_exc_fx, / o : excitation for SWB TBE Q0*/
		@@ -558,7 +562,12 @@ void encod_gen_voic_ivas_fx(
		set16_fx( code_preQ_fx, 0, L_SUBFR );

		shift_wsp = add( Q_new, shift );
		#ifdef CLEANUP_ACELP_ENC
		test();
		if ( LT_32( st_fx->core_brate, MIN_BRATE_AVQ_EXC ) && st_fx->element_mode != EVS_MONO )
		#else
		if ( LT_32( st_fx->core_brate, MIN_BRATE_AVQ_EXC ) )
		#endif
		{
		shift_wsp = sub( shift_wsp, 1 );
		}
		@@ -595,6 +604,20 @@ void encod_gen_voic_ivas_fx(

		Copy( &res_fx[i_subfr_fx], &exc_fx[i_subfr_fx], L_SUBFR ); /Q_new/

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		find_targets_fx( speech_fx, hLPDmem->mem_syn, i_subfr_fx, &hLPDmem->mem_w0, p_Aq_fx,
		res_fx, L_SUBFR, p_Aw_fx, st_fx->preemph_fac, xn_fx, cn_fx, h1_fx );

		Copy_Scale_sig( h1_fx, h2_fx, L_SUBFR, -2 ); /Q11/
		Scale_sig( h1_fx, L_SUBFR, add( 1, shift ) ); /* set h1[] in Q14 with scaling for convolution Q(14+shift)*/
		/* scaling of xn[] to limit dynamic at 12 bits */
		Scale_sig( xn_fx, L_SUBFR, shift );
		}
		ELSE
		#endif
		{
		find_targets_ivas_fx( speech_fx, hLPDmem->mem_syn, i_subfr_fx, &hLPDmem->mem_w0, p_Aq_fx,
		res_fx, L_SUBFR, p_Aw_fx, st_fx->preemph_fac, xn_fx, cn_fx, h1_fx );

		@@ -611,6 +634,7 @@ void encod_gen_voic_ivas_fx(
		rem_offset( xn_fx, xn_fx, L_SUBFR );
		rem_offset( cn_fx, cn_fx, L_SUBFR );
		}
		}

		/----------------------------------------------------------------
		* Close-loop pitch search and quantization
		@@ -630,7 +654,16 @@ void encod_gen_voic_ivas_fx(
		* Find adaptive exitation
		-----------------------------------------------------------------/

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		pred_lt4( &exc_fx[i_subfr_fx], &exc_fx[i_subfr_fx], T0_fx, T0_frac_fx, L_SUBFR + 1, pitch_inter4_2, L_INTERPOL2, PIT_UP_SAMP );
		}
		ELSE
		#endif
		{
		pred_lt4_ivas_fx( &exc_fx[i_subfr_fx], &exc_fx[i_subfr_fx], T0_fx, T0_frac_fx, L_SUBFR + 1, L_pitch_inter4_2, L_INTERPOL2, PIT_UP_SAMP );
		}

		/-----------------------------------------------------------------
		* Gain clipping test to avoid unstable synthesis on frame erasure
		@@ -721,7 +754,14 @@ void encod_gen_voic_ivas_fx(
		gcode16 = round_fx_sat( Lgcode );

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		hLPDmem->tilt_code = Est_tilt2( &exc_fx[i_subfr_fx], gain_pit_fx, code_fx, Lgcode, &voice_fac_fx, shift ); /* Q15 */
		}
		ELSE
		{
		hLPDmem->tilt_code = est_tilt_fx( &exc_fx[i_subfr_fx], gain_pit_fx, code_fx, gain_code_fx, &voice_fac_fx, Q_new, L_SUBFR ); /* Q15 */
		}
		#else
		hLPDmem->tilt_code = est_tilt_ivas_fx( &exc_fx[i_subfr_fx], gain_pit_fx, code_fx, gain_code_fx, &voice_fac_fx, Q_new, L_SUBFR, 0 ); /* Q15 */
		#endif
		@@ -747,6 +787,21 @@ void encod_gen_voic_ivas_fx(
		* Update memory of the weighting filter
		-----------------------------------------------------------------/

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		/* st_fx->mem_w0 = xn[L_SUBFR-1] - (gain_pity1[L_SUBFR-1]) - (gain_codey2[L_SUBFR-1]); */
		Ltmp = L_mult( gcode16, y2_fx[L_SUBFR - 1] ); /Q10/
		Ltmp = L_shl( Ltmp, add( 5, shift ) ); /Q15+shift/
		Ltmp = L_negate( Ltmp );
		Ltmp = L_mac( Ltmp, xn_fx[L_SUBFR - 1], 16384 /Q14/ ); /* Q_new-1+shift+14+1 */
		Ltmp = L_msu( Ltmp, y1_fx[L_SUBFR - 1], gain_pit_fx /Q14/ ); /* Q_new-1+shift+14+1 */
		Ltmp = L_shl_sat( Ltmp, sub( 1, shift ) ); /* 15 + Q_new */
		hLPDmem->mem_w0 = round_fx_sat( Ltmp ); /Q_new-1 /
		}
		ELSE
		#endif
		{
		/*At this point
		xn has to be in Qnew
		yn has to be in Qnew
		@@ -771,6 +826,8 @@ void encod_gen_voic_ivas_fx(
		Ltmp = L_shl( Ltmp, sub( 1, shift ) ); // Q14 + Qnew + 1
		hLPDmem->mem_w0 = round_fx_sat( Ltmp ); /Q_new-1 /
		move16();
		}

		IF( gain_preQ_fx != 0 )
		{
		tmp1_fx = add( 16 - ( 2 + Q_AVQ_OUT_DEC + 1 ), Q_new );
		@@ -787,10 +844,10 @@ void encod_gen_voic_ivas_fx(
		move16();

		/* code in Q9, gain_pit in Q14 */
		Ltmp = L_mult( gcode16, code_fx[i] ); /Qnew + 9 + 1 /
		Ltmp = L_shl( Ltmp, 5 ); /Qnew + 9+ 1+5 /
		Ltmp = L_mac( Ltmp, exc_fx[i + i_subfr_fx], gain_pit_fx ); /Qnew + 14 + 1/
		Ltmp = L_shl_sat( Ltmp, 1 ); /Qnew + 14 + 1 +1 / /* saturation can occur here Q16*/
		Ltmp = L_mult( gcode16, code_fx[i] ); /Q10 / Qnew + 9 + 1 /
		Ltmp = L_shl( Ltmp, 5 ); /Q15 / Qnew + 9+ 1+5 /
		Ltmp = L_mac( Ltmp, exc_fx[i + i_subfr_fx], gain_pit_fx ); /Q15 / Qnew + 14 + 1/
		Ltmp = L_shl_sat( Ltmp, 1 ); /Q15 / Qnew + 14 + 1 +1 /
		exc_fx[i + i_subfr_fx] = round_fx_sat( L_add_sat( Ltmp, Ltmp1 ) );
		move16();
		}
		@@ -806,10 +863,10 @@ void encod_gen_voic_ivas_fx(
		{
		/* code in Q9, gain_pit in Q14 */
		/gcode16 in Qnew/
		Ltmp = L_mult( gcode16, code_fx[i] ); /Qnew + 9 + 1 /
		Ltmp = L_shl_sat( Ltmp, 5 ); /Qnew + 9+ 1+5 /
		Ltmp = L_mac_sat( Ltmp, exc_fx[i + i_subfr_fx], gain_pit_fx ); /Qnew + 14 + 1/
		Ltmp = L_shl_sat( Ltmp, 1 ); /Qnew + 14 + 1 +1 /
		Ltmp = L_mult( gcode16, code_fx[i] ); /Q10 / Qnew + 9 + 1 /
		Ltmp = L_shl_sat( Ltmp, 5 ); /Q15 / Qnew + 9+ 1+5 /
		Ltmp = L_mac_sat( Ltmp, exc_fx[i + i_subfr_fx], gain_pit_fx ); /Q15 / Qnew + 14 + 1/
		Ltmp = L_shl_sat( Ltmp, 1 ); /Q15 / Qnew + 14 + 1 +1 /
		exc_fx[i + i_subfr_fx] = round_fx_sat( Ltmp );
		}
		}
		@@ -828,7 +885,16 @@ void encod_gen_voic_ivas_fx(
		* Update A(z) filters
		-----------------------------------------------------------------/

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		Syn_filt_s( 1, p_Aq_fx, M, &exc_fx[i_subfr_fx], &syn_fx[i_subfr_fx], L_SUBFR, hLPDmem->mem_syn, 1 );
		}
		ELSE
		#endif
		{
		syn_filt_fx( 1, p_Aq_fx, M, &exc_fx[i_subfr_fx], &syn_fx[i_subfr_fx], L_SUBFR, hLPDmem->mem_syn, 1 );
		}

		p_Aw_fx += ( M + 1 );
		p_Aq_fx += ( M + 1 );

lib_enc/enc_pit_exc_fx.c

+14 −0

Original line number	Diff line number	Diff line
		@@ -374,6 +374,14 @@ void enc_pit_exc_fx(
		}
		ELSE IF( use_fcb == 2 )
		{
		#ifdef CLEANUP_ACELP_ENC
		/-----------------------------------------------------------------
		* Innovation encoding
		-----------------------------------------------------------------/

		inov_encode_fx( st_fx, st_fx->core_brate, 0, st_fx->L_frame, st_fx->last_L_frame, GENERIC, st_fx->bwidth, 0, i_subfr, -1, p_Aq,
		gain_pit, cn, exc, h2, st_fx->hLPDmem->tilt_code, pt_pitch, xn2, code, y2, &unbits_PI, 2 L_SUBFR, shift, Q_new );
		#else
		#ifdef FIX_1904_HARM_GSC_ENC
		IF( NE_16( st_fx->element_mode, EVS_MONO ) )
		{
		@@ -400,6 +408,7 @@ void enc_pit_exc_fx(
		gain_pit, cn, exc, h2, st_fx->hLPDmem->tilt_code, pt_pitch, xn2, code, y2, &unbits_PI, 2 L_SUBFR, shift );
		#endif
		}
		#endif

		/-----------------------------------------------------------------
		* Gain encoding
		@@ -413,6 +422,10 @@ void enc_pit_exc_fx(
		* Innovation & gain encoding
		-----------------------------------------------------------------/

		#ifdef CLEANUP_ACELP_ENC
		inov_encode_fx( st_fx, Local_BR, 0, st_fx->L_frame, st_fx->last_L_frame, LOCAL_CT, WB, 1, i_subfr, -1, p_Aq,
		gain_pit, cn, exc, h2, hLPDmem->tilt_code, *pt_pitch, xn2, code, y2, &unbits_PI, L_SUBFR, shift, Q_new );
		#else
		#ifdef FIX_1904_HARM_GSC_ENC
		IF( NE_16( st_fx->element_mode, EVS_MONO ) )
		{
		@@ -435,6 +448,7 @@ void enc_pit_exc_fx(
		gain_pit, cn, exc, h2, hLPDmem->tilt_code, *pt_pitch, xn2, code, y2, &unbits_PI, L_SUBFR, shift );
		#endif
		}
		#endif

		/-----------------------------------------------------------------
		* Gain encoding

lib_enc/inov_enc_fx.c

+4 −1

Original line number	Diff line number	Diff line
		@@ -370,7 +370,7 @@ Word16 inov_encode_ivas_fx(
		Word16 unbits, / o : number of unused bits for PI */
		const Word16 L_subfr, /* i : subframe length */
		const Word16 shift,
		const Word16 Q_new )
		Word16 Q_new )
		{
		Word16 dn[2 * L_SUBFR], Qdn, Qcn, Qh2;
		Word16 nBits, cmpl_flag;
		@@ -422,6 +422,9 @@ Word16 inov_encode_ivas_fx(
		#ifdef CLEANUP_ACELP_ENC
		ELSE
		{
		Q_new = 0;
		move16();

		Qxn = 0;
		move16();
		Qcn = 0;

lib_enc/prot_fx_enc.h

+7 −2

Original line number	Diff line number	Diff line
		@@ -2071,6 +2071,7 @@ ivas_error encod_ppp_ivas_fx(
		Word16 Q_new,
		Word16 shift );

		#ifndef CLEANUP_ACELP_ENC
		void encod_gen_voic_fx(
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx[], /* i : i speech Q0*/
		@@ -2089,8 +2090,12 @@ void encod_gen_voic_fx(
		const Word16 tdm_Pri_pitch_buf[], /* i : primary channel pitch buffer Q6*/
		Word16 shift,
		Word16 Q_new );

		#endif
		#ifdef CLEANUP_ACELP_ENC
		void encod_gen_voic_fx(
		#else
		void encod_gen_voic_ivas_fx(
		#endif
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx[], /* i : input speech Q0*/
		const Word16 Aw_fx[], /* i : weighted A(z) unquantized for subframes Q12*/
		@@ -3564,7 +3569,7 @@ Word16 inov_encode_fx(
		Word16 unbits, / o : number of unused bits for PI */
		const Word16 L_subfr, /* i : subframe length */
		const Word16 shift,
		const Word16 Q_new );
		Word16 Q_new );

		void gain_enc_mless_fx(
		BSTR_ENC_HANDLE hBstr, /* i/o: encoder bitstream handle */