Fix for 3GPP issue 1499: Spike in BASOP encoded signal of MASA selection material (f2d5e62e) · Commits · SA4 / Audio / IVAS BASOP

lib_enc/analy_sp_fx.c

+5 −5

Original line number	Diff line number	Diff line
		@@ -438,7 +438,7 @@ void ivas_analy_sp_fx(
		Word16 q_fr_bands, / o : energy in critical frequency bands Q0 */
		Word32 lf_E, / o : per bin E for first... q_lf_E */
		Word16 q_lf_E, / o : per bin E for first... Q0 */
		Word16 Etot, / o : total input energy Q8 */
		Word32 Etot, / o : total input energy Q24 */
		const Word16 min_band, /* i : minimum critical band Q0 */
		const Word16 max_band, /* i : maximum critical band Q0 */
		Word32 Bin_E, / o : per-bin energy spectrum q_Bin_E */
		@@ -557,7 +557,7 @@ void ivas_analy_sp_fx(

		/* Average total log energy over both half-frames */
		/* Etot = 10.0f (float)log10(0.5f * Etot); /
		Etot = -12800 / 10.f * logf(0.00001f) in Q8 */;
		Etot = -838860800 / 10.f * log10f(0.00001f) in Q24 : This is when LEtot is 0*/;
		move16();
		IF( LEtot != 0 )
		{
		@@ -566,7 +566,7 @@ void ivas_analy_sp_fx(
		LEtot = W_shl( LEtot, exp ); // q_fr_bands+2+exp
		Ltmp = BASOP_Util_Log10( W_extract_h( LEtot ), sub( 61, add( q_fr_bands, exp ) ) / 31-(q_fr_bands+2+exp-32) */ ); // Q25
		Ltmp = Mpy_32_32( Ltmp, 1342177280 /* 10.f in Q27 */ ); // (Q25, Q27) -> Q21
		*Etot = extract_h( L_shl( Ltmp, Q24 - Q21 ) ); // Q8
		*Etot = L_shl( Ltmp, Q24 - Q21 ); // Q24
		move16();
		}
		}
		@@ -599,7 +599,7 @@ void ivas_analy_sp_fx(
		}

		/* Average total log energy over both half-frames */
		Etot = -12800 / 10.f * logf(0.00001f) in Q8 */;
		Etot = -838860800 / 10.f * log10f(0.00001f) in Q24 : This is when LEtot is 0*/;
		move16();
		IF( LEtot != 0 )
		{
		@@ -607,7 +607,7 @@ void ivas_analy_sp_fx(
		LEtot = W_shl( LEtot, exp ); // q_fr_bands+exp
		Ltmp = BASOP_Util_Log10( W_extract_h( LEtot ), sub( 62, add( q_fr_bands, exp ) ) / 31-(q_fr_bands+1+exp-32) */ ); // Q25
		Ltmp = Mpy_32_32( Ltmp, 1342177280 /* 10.f in Q27 */ ); // (Q25, Q27) -> Q21
		*Etot = extract_h( L_shl( Ltmp, Q24 - Q21 ) ); // Q8
		*Etot = L_shl( Ltmp, Q24 - Q21 ); // Q24
		move16();
		}
		}

lib_enc/find_uv_fx.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -729,7 +729,7 @@ Word16 find_uv_ivas_fx( /* o : coding typ
		* Total frame energy difference (dE3)
		-----------------------------------------------------------------/

		dE3 = sub( Etot, hNoiseEst->Etot_last_fx ); /Q8/
		dE3 = sub( Etot, extract_h( hNoiseEst->Etot_last_32fx ) ); /Q8/

		/-----------------------------------------------------------------
		* Energy decrease after spike (dE2)

lib_enc/ivas_core_pre_proc_front_fx.c

+16 −16

Original line number	Diff line number	Diff line
		@@ -205,7 +205,7 @@ ivas_error pre_proc_front_ivas_fx(
		Word16 S_map_fx[L_FFT / 2];
		Word16 cor_map_sum_LR_fx[CPE_CHANNELS]; /* speech/music clasif. parameter */
		Word16 S_map_LR_fx[L_FFT / 2]; /* short-term correlation map */
		Word16 Etot_fx; /* total energy Q8 */
		Word32 Etot_fx; /* total energy Q8 */
		Word32 tmpN_fx[NB_BANDS]; /* Temporary noise update */
		Word32 tmpE_fx[NB_BANDS]; /* Temporary averaged energy of 2 sf. */
		Word32 tmpN_LR_fx[CPE_CHANNELS][NB_BANDS]; /* Temporary noise update */
		@@ -739,7 +739,7 @@ ivas_error pre_proc_front_ivas_fx(

		IF( hStereoClassif != NULL )
		{
		IF( GT_32( sub( st->lp_speech_fx, Etot_fx ), 25 << Q8 ) ) /Q8/
		IF( GT_32( sub( st->lp_speech_fx, extract_h( Etot_fx ) ), 25 << Q8 ) ) /Q8/
		{
		hStereoClassif->silence_flag = 2;
		move16();
		@@ -759,11 +759,11 @@ ivas_error pre_proc_front_ivas_fx(

		IF( hCPE != NULL )
		{
		noise_est_pre_32fx( L_deposit_h( Etot_fx ), st->ini_frame, st->hNoiseEst, st->idchan, element_mode, hCPE->last_element_mode );
		noise_est_pre_32fx( ( Etot_fx ), st->ini_frame, st->hNoiseEst, st->idchan, element_mode, hCPE->last_element_mode );
		}
		ELSE
		{
		noise_est_pre_32fx( L_deposit_h( Etot_fx ), st->ini_frame, st->hNoiseEst, st->idchan, element_mode, element_mode );
		noise_est_pre_32fx( ( Etot_fx ), st->ini_frame, st->hNoiseEst, st->idchan, element_mode, element_mode );
		}

		test();
		@@ -913,18 +913,18 @@ ivas_error pre_proc_front_ivas_fx(
		----------------------------------------------------------------/

		noise_est_down_ivas_fx( fr_bands_fx, fr_bands_fx_q, st->hNoiseEst->bckr_fx, &st->hNoiseEst->q_bckr, tmpN_fx, &q_tmpN, tmpE_fx, &q_tmpE, st->min_band, st->max_band,
		&st->hNoiseEst->totalNoise_fx, Etot_fx, &st->hNoiseEst->Etot_last_fx, &st->hNoiseEst->Etot_v_h2_fx );
		&st->hNoiseEst->totalNoise_fx, Etot_fx, &st->hNoiseEst->Etot_last_32fx, &st->hNoiseEst->Etot_v_h2_fx );

		test();
		IF( lr_vad_enabled && st->idchan == 0 )
		{
		noise_est_down_ivas_fx( fr_bands_LR_fx[0], fr_bands_LR_fx_q[0], hCPE->hFrontVad[0]->hNoiseEst->bckr_fx, &hCPE->hFrontVad[0]->hNoiseEst->q_bckr, tmpN_LR_fx[0], &q_tmpN_LR[0], tmpE_LR_fx[0], &q_tmpE_LR[0], st->min_band, st->max_band, &hCPE->hFrontVad[0]->hNoiseEst->totalNoise_fx, Etot_LR_fx[0], &hCPE->hFrontVad[0]->hNoiseEst->Etot_last_fx, &hCPE->hFrontVad[0]->hNoiseEst->Etot_v_h2_fx );
		noise_est_down_ivas_fx( fr_bands_LR_fx[1], fr_bands_LR_fx_q[1], hCPE->hFrontVad[1]->hNoiseEst->bckr_fx, &hCPE->hFrontVad[1]->hNoiseEst->q_bckr, tmpN_LR_fx[1], &q_tmpN_LR[1], tmpE_LR_fx[1], &q_tmpE_LR[1], st->min_band, st->max_band, &hCPE->hFrontVad[1]->hNoiseEst->totalNoise_fx, Etot_LR_fx[1], &hCPE->hFrontVad[1]->hNoiseEst->Etot_last_fx, &hCPE->hFrontVad[1]->hNoiseEst->Etot_v_h2_fx );
		noise_est_down_ivas_fx( fr_bands_LR_fx[0], fr_bands_LR_fx_q[0], hCPE->hFrontVad[0]->hNoiseEst->bckr_fx, &hCPE->hFrontVad[0]->hNoiseEst->q_bckr, tmpN_LR_fx[0], &q_tmpN_LR[0], tmpE_LR_fx[0], &q_tmpE_LR[0], st->min_band, st->max_band, &hCPE->hFrontVad[0]->hNoiseEst->totalNoise_fx, L_deposit_h( Etot_LR_fx[0] ), &hCPE->hFrontVad[0]->hNoiseEst->Etot_last_32fx, &hCPE->hFrontVad[0]->hNoiseEst->Etot_v_h2_fx );
		noise_est_down_ivas_fx( fr_bands_LR_fx[1], fr_bands_LR_fx_q[1], hCPE->hFrontVad[1]->hNoiseEst->bckr_fx, &hCPE->hFrontVad[1]->hNoiseEst->q_bckr, tmpN_LR_fx[1], &q_tmpN_LR[1], tmpE_LR_fx[1], &q_tmpE_LR[1], st->min_band, st->max_band, &hCPE->hFrontVad[1]->hNoiseEst->totalNoise_fx, L_deposit_h( Etot_LR_fx[1] ), &hCPE->hFrontVad[1]->hNoiseEst->Etot_last_32fx, &hCPE->hFrontVad[1]->hNoiseEst->Etot_v_h2_fx );

		corr_shiftL_fx = correlation_shift_fx( hCPE->hFrontVad[0]->hNoiseEst->totalNoise_fx ); // Q15
		corr_shiftR_fx = correlation_shift_fx( hCPE->hFrontVad[1]->hNoiseEst->totalNoise_fx ); // Q15
		}
		*relE_fx = sub( Etot_fx, st->lp_speech_fx ); // Q8
		*relE_fx = sub( extract_h( Etot_fx ), st->lp_speech_fx ); // Q8
		move16();

		corr_shift_fx = correlation_shift_fx( st->hNoiseEst->totalNoise_fx ); /* Q15 */
		@@ -1307,8 +1307,8 @@ ivas_error pre_proc_front_ivas_fx(
		move16();
		move16();

		noise_est_ivas_fx( st, old_pitch1, tmpN_fx, epsP_fx, Etot_fx, *relE_fx, corr_shift_fx, tmpE_fx, q_tmpE, fr_bands_fx, fr_bands_fx_q, cor_map_sum_fx,
		&ncharX_fx, &sp_div_fx, &q_sp_div, &non_staX_fx, loc_harm, lf_E_fx, q_lf_E_fx, &st->hNoiseEst->harm_cor_cnt, st->hNoiseEst->Etot_l_lp_fx,
		noise_est_ivas_fx( st, old_pitch1, tmpN_fx, epsP_fx, extract_h( Etot_fx ), *relE_fx, corr_shift_fx, tmpE_fx, q_tmpE, fr_bands_fx, fr_bands_fx_q, cor_map_sum_fx,
		&ncharX_fx, &sp_div_fx, &q_sp_div, &non_staX_fx, loc_harm, lf_E_fx, q_lf_E_fx, &st->hNoiseEst->harm_cor_cnt, extract_h( st->hNoiseEst->Etot_l_lp_32fx ),
		st->hNoiseEst->Etot_v_h2_fx, &st->hNoiseEst->bg_cnt, st->lgBin_E_fx, &dummy_fx, S_map_fx,
		hStereoClassif, NULL, st->ini_frame );

		@@ -1352,14 +1352,14 @@ ivas_error pre_proc_front_ivas_fx(
		noise_est_ivas_fx( st, old_pitch1, tmpN_LR_fx[0], epsP_fx, Etot_LR_fx[0], sub( Etot_LR_fx[0], hCPE->hFrontVad[0]->lp_speech_fx ), corr_shiftL_fx,
		tmpE_LR_fx[0], q_tmpE_LR[0], fr_bands_LR_fx[0], fr_bands_LR_fx_q[0], &cor_map_sum_LR_fx[0], &ncharX_LR_fx, &sp_div_LR_fx, &q_sp_div_LR,
		&non_staX_LR_fx, loc_harmLR_fx, lf_E_LR_fx[0], lf_E_LR_fx_q, &hCPE->hFrontVad[0]->hNoiseEst->harm_cor_cnt,
		hCPE->hFrontVad[0]->hNoiseEst->Etot_l_lp_fx, hCPE->hFrontVad[0]->hNoiseEst->Etot_v_h2_fx, &hCPE->hFrontVad[0]->hNoiseEst->bg_cnt,
		extract_h( hCPE->hFrontVad[0]->hNoiseEst->Etot_l_lp_32fx ), hCPE->hFrontVad[0]->hNoiseEst->Etot_v_h2_fx, &hCPE->hFrontVad[0]->hNoiseEst->bg_cnt,
		st->lgBin_E_fx, &dummy_fx, S_map_LR_fx, NULL, hCPE->hFrontVad[0], hCPE->hFrontVad[0]->ini_frame );

		/* Note: the index [0] in the last argument is intended, the ini_frame counter is only maintained in the zero-th channel's VAD handle */
		noise_est_ivas_fx( st, old_pitch1, tmpN_LR_fx[1], epsP_fx, Etot_LR_fx[1], sub( Etot_LR_fx[1], hCPE->hFrontVad[1]->lp_speech_fx ), corr_shiftR_fx,
		tmpE_LR_fx[1], q_tmpE_LR[1], fr_bands_LR_fx[1], fr_bands_LR_fx_q[1], &cor_map_sum_LR_fx[1], &ncharX_LR_fx, &sp_div_LR_fx, &q_sp_div_LR,
		&non_staX_LR_fx, loc_harmLR_fx, lf_E_LR_fx[1], lf_E_LR_fx_q, &hCPE->hFrontVad[1]->hNoiseEst->harm_cor_cnt,
		hCPE->hFrontVad[1]->hNoiseEst->Etot_l_lp_fx, hCPE->hFrontVad[1]->hNoiseEst->Etot_v_h2_fx, &hCPE->hFrontVad[1]->hNoiseEst->bg_cnt,
		extract_h( hCPE->hFrontVad[1]->hNoiseEst->Etot_l_lp_32fx ), hCPE->hFrontVad[1]->hNoiseEst->Etot_v_h2_fx, &hCPE->hFrontVad[1]->hNoiseEst->bg_cnt,
		st->lgBin_E_fx, &dummy_fx, S_map_LR_fx, NULL, hCPE->hFrontVad[1], hCPE->hFrontVad[0]->ini_frame );
		}

		@@ -1387,7 +1387,7 @@ ivas_error pre_proc_front_ivas_fx(
		find_tilt_ivas_fx( fr_bands_fx, fr_bands_fx_q, st->hNoiseEst->bckr_fx, st->hNoiseEst->q_bckr, ee_fx, st->pitch, st->voicing_fx, lf_E_fx, q_lf_E_fx,
		corr_shift_fx, st->input_bwidth, st->max_band, hp_E_fx, MODE1, &( st->bckr_tilt_lt ), st->Opt_SC_VBR );

		st->coder_type = find_uv_ivas_fx( st, pitch_fr_fx, voicing_fr_fx, inp_12k8_fx, ee_fx, &dE1X_fx, corr_shift_fx, *relE_fx, Etot_fx, hp_E_fx,
		st->coder_type = find_uv_ivas_fx( st, pitch_fr_fx, voicing_fr_fx, inp_12k8_fx, ee_fx, &dE1X_fx, corr_shift_fx, *relE_fx, extract_h( Etot_fx ), hp_E_fx,
		&flag_spitch, last_core_orig, hStereoClassif, Q_new /q_inp_12k8*/, fr_bands_fx_q ); // Q0
		Copy_Scale_sig_16_32_no_sat( st->lgBin_E_fx, st->Bin_E_fx, L_FFT / 2, sub( st->q_Bin_E, Q7 ) );

		@@ -1461,7 +1461,7 @@ ivas_error pre_proc_front_ivas_fx(
		scale_sig32( PS_fx, 128, shift );
		Qfact_PS = add( Qfact_PS, shift );

		smc_dec = ivas_smc_gmm_fx( st, hStereoClassif, localVAD_HE_SAD, Etot_fx, lsp_new_fx, cor_map_sum_fx /Q8/, epsP_fx, PS_fx, non_staX_fx, relE_fx, &high_lpn_flag, flag_spitch, Qfact_PS, epsP_fx_q, hSpMusClas->past_PS_Q ); / Q0 */
		smc_dec = ivas_smc_gmm_fx( st, hStereoClassif, localVAD_HE_SAD, extract_h( Etot_fx ), lsp_new_fx, cor_map_sum_fx /Q8/, epsP_fx, PS_fx, non_staX_fx, relE_fx, &high_lpn_flag, flag_spitch, Qfact_PS, epsP_fx_q, hSpMusClas->past_PS_Q ); / Q0 */

		#ifdef DEBUG_FORCE_DIR
		if ( st->force_dir[0] != '\0' )
		@@ -1495,7 +1495,7 @@ ivas_error pre_proc_front_ivas_fx(
		* Update of old per-band energy spectrum
		----------------------------------------------------------------/

		ivas_long_enr_fx( st, Etot_fx, localVAD_HE_SAD, high_lpn_flag, NULL, 1, NULL, NULL );
		ivas_long_enr_fx( st, extract_h( Etot_fx ), localVAD_HE_SAD, high_lpn_flag, NULL, 1, NULL, NULL );

		Copy32( fr_bands_fx + NB_BANDS, st->hNoiseEst->enrO_fx, NB_BANDS ); /* fr_bands_fx_q */
		st->hNoiseEst->q_enrO = fr_bands_fx_q;
		@@ -1676,7 +1676,7 @@ ivas_error pre_proc_front_ivas_fx(
		st->q_Bin_E = add( *Q_new, Q_SCALE - 2 );
		move16();
		/* 2nd stage speech/music classification (ACELP/GSC/TCX core selection) */
		ivas_smc_mode_selection_fx( st, element_brate, smc_dec, relE_fx, Etot_fx, attack_flag, inp_12k8_fx, Q_new, S_map_fx, flag_spitch );
		ivas_smc_mode_selection_fx( st, element_brate, smc_dec, relE_fx, extract_h( Etot_fx ), attack_flag, inp_12k8_fx, Q_new, S_map_fx, flag_spitch );
		}

		#ifdef DEBUG_FORCE_DIR

lib_enc/ivas_cpe_enc_fx.c

+2 −2

Original line number	Diff line number	Diff line
		@@ -735,8 +735,8 @@ ivas_error ivas_cpe_enc_fx(
		move16();
		}

		Etot_last_fx[0] = sts[0]->hNoiseEst->Etot_last_fx; /* Q8 */
		Etot_last_fx[1] = sts[1]->hNoiseEst->Etot_last_fx; /* Q8 */
		Etot_last_fx[0] = extract_h( sts[0]->hNoiseEst->Etot_last_32fx ); /* Q8 */
		Etot_last_fx[1] = extract_h( sts[1]->hNoiseEst->Etot_last_32fx ); /* Q8 */
		move16();
		move16();
		}

lib_enc/ivas_front_vad_fx.c

+7 −5

Original line number	Diff line number	Diff line
		@@ -90,6 +90,7 @@ ivas_error front_vad_fx(
		ivas_error error;
		Word16 Q_new;
		Word16 Qband, mem_decim_size;
		Word32 Etot_LR_32fx;
		error = IVAS_ERR_OK;
		push_wmops( "front_vad" );
		move16();
		@@ -248,7 +249,7 @@ ivas_error front_vad_fx(
		move16();

		ivas_analy_sp_fx( IVAS_CPE_TD, hCPE, sts[0]->input_Fs, hFrontVad->buffer_12k8_fx + L_FFT / 2 - 3 * ( L_SUBFR / 2 ), Q_new, fr_bands_fx[n],
		&q_fr_bands[n], lf_E_fx[n], &q_lf_E[n], &Etot_LR_fx[n], sts[0]->min_band, sts[0]->max_band, Bin_E_fx, &q_Bin_E, Bin_E_old_fx,
		&q_fr_bands[n], lf_E_fx[n], &q_lf_E[n], &Etot_LR_32fx, sts[0]->min_band, sts[0]->max_band, Bin_E_fx, &q_Bin_E, Bin_E_old_fx,
		&q_Bin_E_old, PS_fx, q_PS_out, lgBin_E_fx, band_energies_fx, &q_band_energies, fft_buffLR_fx, &q_fft_buffLR );
		if ( n == 0 )
		{
		@@ -275,9 +276,10 @@ ivas_error front_vad_fx(

		/* add up energies for later calculating average of channel energies */

		Word32 Etot_fx = L_deposit_h( Etot_LR_fx[n] ); /* Q24 */
		noise_est_pre_32fx( Etot_LR_32fx, hFrontVads[0]->ini_frame, hFrontVad->hNoiseEst, 0, 0, 0 );

		noise_est_pre_32fx( Etot_fx, hFrontVads[0]->ini_frame, hFrontVad->hNoiseEst, 0, 0, 0 );
		Etot_LR_fx[n] = extract_h( Etot_LR_32fx );
		move16();

		/* wb_vad */
		Word16 scale = s_min( q_fr_bands[n], add( hFrontVads[n]->hNoiseEst->q_enrO, L_norm_arr( hFrontVads[n]->hNoiseEst->enrO_fx, NB_BANDS ) ) );
		@@ -611,7 +613,7 @@ ivas_error front_vad_spar_fx(

		Word16 q_tmpN, q_tmpE;
		noise_est_down_ivas_fx( fr_bands_fx[0], q_fr_bands[0], hFrontVad->hNoiseEst->bckr_fx, &hFrontVad->hNoiseEst->q_bckr, tmpN_fx, &q_tmpN, tmpE_fx, &q_tmpE, st->min_band, st->max_band,
		&hFrontVad->hNoiseEst->totalNoise_fx, Etot_fx[0], &hFrontVad->hNoiseEst->Etot_last_fx, &hFrontVad->hNoiseEst->Etot_v_h2_fx );
		&hFrontVad->hNoiseEst->totalNoise_fx, Etot_fx[0], &hFrontVad->hNoiseEst->Etot_last_32fx, &hFrontVad->hNoiseEst->Etot_v_h2_fx );

		corr_shift_fx = correlation_shift_fx( hFrontVad->hNoiseEst->totalNoise_fx ); /* Q15 */

		@@ -753,7 +755,7 @@ ivas_error front_vad_spar_fx(

		noise_est_ivas_fx( st, old_pitch, tmpN_fx, epsP_fx, Etot_fx[0], sub( Etot_fx[0], hFrontVad->lp_speech_fx ), corr_shift_fx, tmpE_fx,
		hFrontVad->hNoiseEst->ave_enr_q, fr_bands_fx[0], q_fr_bands[0], &cor_map_sum_fx, NULL, &sp_div_fx, &Q_sp_div, &non_staX_fx, &loc_harm,
		lf_E_fx[0], q_lf_E[0], &hFrontVad->hNoiseEst->harm_cor_cnt, hFrontVad->hNoiseEst->Etot_l_lp_fx, hFrontVad->hNoiseEst->Etot_v_h2_fx,
		lf_E_fx[0], q_lf_E[0], &hFrontVad->hNoiseEst->harm_cor_cnt, extract_h( hFrontVad->hNoiseEst->Etot_l_lp_32fx ), hFrontVad->hNoiseEst->Etot_v_h2_fx,
		&hFrontVad->hNoiseEst->bg_cnt, st->lgBin_E_fx, &sp_floor, S_map_fx, NULL, hFrontVad, hFrontVad->ini_frame );

		MVR2R_WORD16( st->pitch, st->pitch, 3 );