harmonize encod_unvoiced_fx(), under CLEANUP_ACELP_ENC (a0501b1d) · Commits · SA4 / Audio / IVAS BASOP

lib_enc/acelp_core_enc_fx.c

+20 −1

Original line number	Diff line number	Diff line
		@@ -95,12 +95,17 @@ ivas_error acelp_core_enc_fx(
		ivas_error error;

		error = IVAS_ERR_OK;

		#ifdef CLEANUP_ACELP_ENC
		(void) hStereoTD;
		#else
		test();
		IF( EQ_16( st_fx->element_mode, IVAS_CPE_MDCT ) && LE_32( st_fx->core_brate, SID_2k40 ) )
		{
		/* Core was ACELP because of DTX in MDCT-Stereo, but SID encoding for that is done in separate function */
		return error;
		}
		#endif
		/------------------------------------------------------------------
		* Initialization
		------------------------------------------------------------------/
		@@ -196,6 +201,7 @@ ivas_error acelp_core_enc_fx(
		move16();
		}

		#ifndef CLEANUP_ACELP_ENC
		/* TD stereo */
		test();
		IF( EQ_16( st_fx->element_mode, IVAS_CPE_TD ) && EQ_16( st_fx->idchan, 1 ) )
		@@ -206,21 +212,25 @@ ivas_error acelp_core_enc_fx(
		tdm_Pri_pitch_buf = hStereoTD->tdm_Pri_pitch_buf_fx;
		}
		ELSE
		#endif
		{
		tdm_lp_reuse_flag = 0;
		tdm_low_rate_mode = 0;
		move16();
		move16();
		#ifndef CLEANUP_ACELP_ENC
		if ( EQ_16( st_fx->element_mode, IVAS_SCE ) && st_fx->low_rate_mode )
		{
		tdm_low_rate_mode = 1;
		move16();
		}
		#endif
		tdm_Pitch_reuse_flag = 0;
		tdm_Pri_pitch_buf = NULL;
		move16();
		move16();
		}

		/-----------------------------------------------------------------
		* ACELP@12k8 / ACELP@16k switching
		-----------------------------------------------------------------/
		@@ -551,21 +561,26 @@ ivas_error acelp_core_enc_fx(
		E_UTIL_deemph2( sub( Q_new, 1 ), syn1_fx, st_fx->preemph_fac, st_fx->L_frame, &hLPDmem->syn[M] );
		Copy( syn1_fx + st_fx->L_frame - M - 1, hLPDmem->syn, M + 1 ); // Q_syn

		#ifndef CLEANUP_ACELP_ENC
		test();
		IF( st_fx->element_mode > EVS_MONO && st_fx->hTcxEnc != NULL )
		{
		Copy( syn1_fx + st_fx->L_frame / 2, st_fx->hTcxEnc->Txnq, st_fx->L_frame / 2 ); // Q_syn
		}

		#endif
		/--------------------------------------------------------------------------------------
		* Modify the excitation signal when the noise is stationary
		--------------------------------------------------------------------------------------/

		#ifdef CLEANUP_ACELP_ENC
		IF( NE_16( nelp_mode, 1 ) )
		#else
		test();
		test();
		test();
		test();
		IF( !( EQ_16( st_fx->idchan, 1 ) && EQ_16( st_fx->element_mode, IVAS_CPE_TD ) ) && NE_16( nelp_mode, 1 ) && !( EQ_16( st_fx->element_mode, IVAS_SCE ) && tdm_low_rate_mode ) )
		#endif
		{
		L_epsP[0] = L_Comp( epsP_h_fx[2], epsP_l_fx[2] );
		move32();
		@@ -1335,7 +1350,11 @@ ivas_error acelp_core_enc_ivas_fx(
		ELSE IF( EQ_16( st->coder_type, UNVOICED ) )
		{
		/* UNVOICED frames (Gauss. excitation) */
		#ifdef CLEANUP_ACELP_ENC
		encod_unvoiced_fx( st, inp, Aw, Aq, Es_pred_fx, uc_two_stage_flag, res_fx, syn_fx, &tmp_noise_fx, exc_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, Q_new, 0 );
		#else
		encod_unvoiced_ivas_fx( st, inp, Aw, Aq, Es_pred_fx, uc_two_stage_flag, res_fx, syn_fx, &tmp_noise_fx, exc_fx, pitch_buf, voice_factors_fx, bwe_exc_fx, Q_new, 0 );
		#endif
		}
		ELSE IF( EQ_16( st->coder_type, TRANSITION ) )
		{

lib_enc/enc_uv_fx.c

+55 −14

Original line number	Diff line number	Diff line
		@@ -16,7 +16,7 @@
		*
		* Encode unvoiced (UC) frames
		-------------------------------------------------------------------/

		#ifndef CLEANUP_ACELP_ENC
		void encod_unvoiced_fx(
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx, / i : Input speech Q_new*/
		@@ -247,9 +247,13 @@ void encod_unvoiced_fx(

		return;
		}
		#endif


		#ifdef CLEANUP_ACELP_ENC
		void encod_unvoiced_fx(
		#else
		void encod_unvoiced_ivas_fx(
		#endif
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx, / i : Input speech Q_new*/
		const Word16 Aw_fx[], /* i : weighted A(z) unquantized for subframes Q12*/
		@@ -257,7 +261,7 @@ void encod_unvoiced_ivas_fx(
		const Word16 Es_pred, /* i : predicted scaled innov. energy Q8*/
		const Word16 uc_two_stage_flag, /* i : flag indicating two-stage UC Q0*/
		const Word16 res_fx, / i : residual signal Q_new*/
		Word16 syn_fx, / o : core synthesis Q_new*/
		Word16 syn_fx, / o : core synthesis Q_new-1 (EVS) / Qnew*/
		Word16 tmp_noise_fx, / o : long-term noise energy Q0*/
		Word16 exc_fx, / i/o: current non-enhanced excitation Q_new*/
		Word16 pitch_buf_fx, / o : floating pitch values for each subframe Q6*/
		@@ -280,7 +284,7 @@ void encod_unvoiced_ivas_fx(
		Word16 y1[L_SUBFR]; /* Filtered adaptive excitation */
		Word16 code2[L_SUBFR]; /* Gaussian excitation */
		Word16 y22[L_SUBFR]; /* Filtered Gaussian excitation */
		// Word16 prm_t[2 * NPRM_DIV], *prm = prm_t;
		Word16 prm_t[2 * NPRM_DIV], *prm = prm_t;
		const Word16 p_Aw_fx, p_Aq_fx; /* pointer to LP filter coeff. vector */
		Word32 norm_gain_code_fx;
		ACELP_config *acelp_cfg;
		@@ -329,12 +333,25 @@ void encod_unvoiced_ivas_fx(
		i_subfr_idx = shr( i_subfr, 6 );
		Copy( &res_fx[i_subfr], &exc_fx[i_subfr], 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, &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);/
		Scale_sig( h1_fx, L_SUBFR, add( 1, shift ) ); /* set h1[] in Q14 with scaling for convolution */
		}
		ELSE
		#endif
		{
		find_targets_ivas_fx( speech_fx, hLPDmem->mem_syn, i_subfr, &hLPDmem->mem_w0, p_Aq_fx,
		res_fx, L_SUBFR, p_Aw_fx, st_fx->preemph_fac, xn_fx, cn_fx, h1_fx );

		q_h1 = sub( 14, norm_s( h1_fx[0] ) );
		Copy_Scale_sig( h1_fx, h2_fx, L_SUBFR, sub( 11, q_h1 ) );
		Scale_sig( h1_fx, L_SUBFR, sub( 14, q_h1 ) );
		}

		/* scaling of xn[] to limit dynamic at 12 bits */
		Scale_sig( xn_fx, L_SUBFR, shift ); // Q_new - 1 + shift
		@@ -366,16 +383,31 @@ void encod_unvoiced_ivas_fx(
		-----------------------------------------------------------------/

		Mode2_gp_clip_fx( st_fx->voicing_fx, i_subfr, st_fx->coder_type, xn_fx, st_fx->clip_var_fx, L_SUBFR, Q_xn );

		*pt_pitch_fx = L_SUBFR << 6;
		move16();
		#ifdef CLEANUP_ACELP_ENC
		if ( st_fx->element_mode == EVS_MONO )
		{
		*pt_pitch_fx = L_SUBFR;
		move16();
		}
		#endif

		/----------------------------------------------------------------------
		* Encode the algebraic innovation *
		----------------------------------------------------------------------/

		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		E_ACELP_innovative_codebook_fx( exc_fx, *pt_pitch_fx, 0, 1, gain_pit_fx, hLPDmem->tilt_code, acelp_cfg, i_subfr, p_Aq_fx, h1_fx, xn_fx, cn_fx, y1, y2_fx, (Word8) st_fx->acelp_autocorr, &prm, code_fx, shift, st_fx->L_frame, st_fx->last_L_frame, st_fx->total_brate );
		}
		ELSE
		{
		inov_encode_fx( st_fx, st_fx->core_brate, 0, L_FRAME, st_fx->last_L_frame, UNVOICED, st_fx->bwidth, 1, i_subfr, -1, p_Aq_fx,
		gain_pit_fx, cn_fx, exc_fx, h2_fx, hLPDmem->tilt_code, *pt_pitch_fx, xn_fx, code_fx, y2_fx, &unbits_PI, L_SUBFR, shift, Q_new );
		}
		#else
		inov_encode_ivas_fx( st_fx, st_fx->core_brate, 0, L_FRAME, st_fx->last_L_frame,
		UNVOICED, st_fx->bwidth, 1, i_subfr, -1, p_Aq_fx,
		@@ -394,7 +426,16 @@ void encod_unvoiced_ivas_fx(
		move16();

		assert( gain_pit_fx == 0 );
		#ifdef CLEANUP_ACELP_ENC
		IF( st_fx->element_mode == EVS_MONO )
		{
		gauss_L2_fx( h1_fx, code2, y2_fx, y22, &gain_code2, &g_corr, gain_pit_fx, hLPDmem->tilt_code, p_Aq_fx, acelp_cfg->formant_enh_num, &( st_fx->seed_acelp ), shift );
		}
		ELSE
		#endif
		{
		gauss_L2_ivas_fx( h1_fx, code2, y2_fx, y22, &gain_code2, &g_corr, gain_pit_fx, hLPDmem->tilt_code, p_Aq_fx, acelp_cfg->formant_enh_num, &( st_fx->seed_acelp ), shift );
		}

		/----------------------------------------------------------
		* - Compute the fixed codebook gain *

lib_enc/prot_fx_enc.h

+2 −2

Original line number	Diff line number	Diff line
		@@ -2010,7 +2010,7 @@ void encod_unvoiced_fx(
		Word16 bwe_exc_fx, / i/o: excitation for SWB TBE Q_exc*/
		const Word16 Q_new,
		const Word16 shift );

		#ifndef CLEANUP_ACELP_ENC
		void encod_unvoiced_ivas_fx(
		Encoder_State st_fx, / i/o: state structure */
		const Word16 speech_fx, / i : Input speech Q_new*/
		@@ -2027,7 +2027,7 @@ void encod_unvoiced_ivas_fx(
		Word16 bwe_exc_fx, / i/o: excitation for SWB TBE Q_exc*/
		const Word16 Q_new,
		const Word16 shift );

		#endif
		void enc_acelp_tcx_main_fx(
		const Word16 new_samples[], /* i : new samples Q15*/
		Encoder_State st, / i/o: encoder state structure */