fix for issue 91: Improvements to SBA high bitrate HOA3 coding (bfe8f7da) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_prot.h

+16 −0

Original line number	Diff line number	Diff line
		@@ -2999,6 +2999,10 @@ void ivas_dirac_param_est_enc(
		float **pp_fr_real,
		float **pp_fr_imag,
		const int16_t input_frame
		#ifdef SBA_HOA_HBR_IMPROV
		,
		const int16_t merged_sba_mode
		#endif
		);

		/----------------------------------------------------------------------------------
		@@ -3866,6 +3870,10 @@ ivas_error ivas_spar_md_enc_open(
		ivas_spar_md_enc_state_t *hMdEnc, / i/o: SPAR MD encoder handle */
		const ENCODER_CONFIG_HANDLE hEncoderConfig, /* i : configuration structure */
		const int16_t sba_order /* i : Ambisonic (SBA) order */
		#ifdef SBA_HOA_HBR_IMPROV
		,
		const int16_t spar_hoa_md_flag /* i : flag to send HOA MD for all bands */
		#endif
		);

		void ivas_spar_md_enc_close(
		@@ -3941,6 +3949,10 @@ ivas_error ivas_spar_md_dec_open(
		ivas_spar_md_dec_state_t *hMdDec_out, / i/o: SPAR MD decoder handle */
		const DECODER_CONFIG_HANDLE hDecoderConfig, /* i : configuration structure */
		const int16_t num_channels /* i : number of internal channels */
		#ifdef SBA_HOA_HBR_IMPROV
		,
		const int16_t spar_hoa_md_flag /* i : flag to send HOA MD for all bands */
		#endif
		);

		void ivas_spar_md_dec_close(
		@@ -4766,6 +4778,10 @@ void computeReferencePower_enc(
		float reference_power, / o : Estimated power */
		const int16_t enc_param_start_band, /* i : first band to process */
		const int16_t num_freq_bands /* i : Number of frequency bands */
		#ifdef SBA_HOA_HBR_IMPROV
		,
		const int16_t merged_sba_mode
		#endif
		);

lib_com/ivas_spar_com.c

+2 −1

Original line number	Diff line number	Diff line
		@@ -1731,10 +1731,11 @@ void ivas_get_spar_md_from_dirac(

		/SPAR from DirAC/
		set_f( response_avg, 0.0f, MAX_OUTPUT_CHANNELS );
		#ifndef SBA_HOA_HBR_IMPROV
		set_f( hSpar_md->band_coeffs[band + i_ts * IVAS_MAX_NUM_BANDS].pred_re, 0.0f, IVAS_SPAR_MAX_CH - 1 );
		set_f( &hSpar_md->band_coeffs[band + i_ts * IVAS_MAX_NUM_BANDS].C_re[0][0], 0.0f, ( IVAS_SPAR_MAX_CH - IVAS_SPAR_MAX_DMX_CHS ) * ( IVAS_SPAR_MAX_DMX_CHS - 1 ) );
		set_f( &hSpar_md->band_coeffs[band + i_ts * IVAS_MAX_NUM_BANDS].P_re[0], 0.0f, ( IVAS_SPAR_MAX_CH - 1 ) );

		#endif
		if ( n_ts > 1 )
		{
		ivas_dirac_dec_get_response( (int16_t) azi_dirac[band][i_ts], (int16_t) ele_dirac[band][i_ts], response_avg, order );

lib_com/options.h

+2 −2

Original line number	Diff line number	Diff line
		@@ -148,7 +148,7 @@

		#define FIX_ADAP_STEFI_SHIFT /* Issue 89: fix bug in parameter shift of adaptive stereo filling */
		#define SPAR_SCALING_HARMONIZATION /* issue 80: Changes to harmonize scaling in spar */

		#define SBA_HOA_HBR_IMPROV /* issue 91: Improvements to SBA high bitrate HOA3 coding */

		/* ################## End DEVELOPMENT switches ######################### */
		/* clang-format on */

lib_dec/ivas_spar_decoder.c

+35 −2

Original line number	Diff line number	Diff line
		@@ -69,6 +69,9 @@ ivas_error ivas_spar_dec_open(
		IVAS_FB_CFG *fb_cfg;
		int16_t i, j, b, active_w_mixing;
		int32_t output_Fs;
		#ifdef SBA_HOA_HBR_IMPROV
		int16_t spar_hoa_md_flag;
		#endif

		error = IVAS_ERR_OK;
		sba_order_internal = min( st_ivas->sba_analysis_order, IVAS_MAX_SBA_ORDER );
		@@ -80,6 +83,17 @@ ivas_error ivas_spar_dec_open(
		return IVAS_ERROR( IVAS_ERR_FAILED_ALLOC, "Can not allocate memory for SPAR decoder" );
		}

		#ifdef SBA_HOA_HBR_IMPROV
		if ( st_ivas->sba_order > 1 && st_ivas->hDecoderConfig->ivas_total_brate >= IVAS_256k )
		{
		spar_hoa_md_flag = 1;
		}
		else
		{
		spar_hoa_md_flag = 0;
		}
		#endif

		output_Fs = st_ivas->hDecoderConfig->output_Fs;

		/* TD decorr. */
		@@ -89,7 +103,12 @@ ivas_error ivas_spar_dec_open(
		}

		/* MD handle */
		if ( ( error = ivas_spar_md_dec_open( &hSpar->hMdDec, st_ivas->hDecoderConfig, num_channels_internal ) ) != IVAS_ERR_OK )
		if ( ( error = ivas_spar_md_dec_open( &hSpar->hMdDec, st_ivas->hDecoderConfig, num_channels_internal
		#ifdef SBA_HOA_HBR_IMPROV
		,
		spar_hoa_md_flag
		#endif
		) ) != IVAS_ERR_OK )
		{
		return error;
		}
		@@ -632,7 +651,16 @@ static void ivas_spar_dec_MD(
		{
		ivas_parse_spar_header( hDecoderConfig->ivas_total_brate, sba_order, st0, &table_idx );

		#ifdef SBA_HOA_HBR_IMPROV
		if ( hSpar->hMdDec->spar_hoa_md_flag )
		{
		hSpar->hMdDec->spar_md.num_bands = IVAS_MAX_NUM_BANDS;
		}
		else
		#endif
		{
		hSpar->hMdDec->spar_md.num_bands = min( SPAR_DIRAC_SPLIT_START_BAND, IVAS_MAX_NUM_BANDS );
		}

		if ( hSpar->hMdDec->table_idx != table_idx )
		{
		@@ -775,7 +803,12 @@ void ivas_spar_get_parameters(

		weight = ivas_spar_get_cldfb_slot_gain( hSpar, hDecoderConfig, ts, &ts0, &ts1, &weight_20ms );

		#ifdef SBA_HOA_HBR_IMPROV
		split_band = hSpar->hMdDec->spar_md.num_bands;
		#else
		split_band = SPAR_DIRAC_SPLIT_START_BAND;
		#endif

		for ( spar_band = 0; spar_band < num_spar_bands; spar_band++ )
		{
		for ( out_ch = 0; out_ch < num_ch_out; out_ch++ )

lib_dec/ivas_spar_md_dec.c

+114 −10

Original line number	Diff line number	Diff line
		@@ -252,6 +252,10 @@ ivas_error ivas_spar_md_dec_open(
		ivas_spar_md_dec_state_t *hMdDec_out, / i/o: SPAR MD decoder handle */
		const DECODER_CONFIG_HANDLE hDecoderConfig, /* i : configuration structure */
		const int16_t num_channels /* i : number of internal channels */
		#ifdef SBA_HOA_HBR_IMPROV
		,
		const int16_t spar_hoa_md_flag /* i : flag to send HOA MD for all bands */
		#endif
		)
		{
		ivas_spar_md_dec_state_t *hMdDec;
		@@ -269,6 +273,9 @@ ivas_error ivas_spar_md_dec_open(
		return error;
		}

		#ifdef SBA_HOA_HBR_IMPROV
		hMdDec->spar_hoa_md_flag = spar_hoa_md_flag;
		#endif
		hMdDec->table_idx = 0; /* just to initialize state variables*/

		if ( ( error = ivas_spar_md_dec_init( hMdDec, hDecoderConfig, num_channels ) ) != IVAS_ERR_OK )
		@@ -440,7 +447,13 @@ ivas_error ivas_spar_md_dec_init(
		float pFC[IVAS_MAX_NUM_BANDS], PR_minmax[2];

		hMdDec->spar_md_cfg.gen_bs = 1; // VE2DB : always 1 - can it be removed?
		ivas_spar_set_bitrate_config( &hMdDec->spar_md_cfg, hMdDec->table_idx, min( IVAS_MAX_NUM_BANDS, SPAR_DIRAC_SPLIT_START_BAND ) );
		ivas_spar_set_bitrate_config( &hMdDec->spar_md_cfg, hMdDec->table_idx,
		#ifdef SBA_HOA_HBR_IMPROV
		( hMdDec->spar_hoa_md_flag ) ? IVAS_MAX_NUM_BANDS : min( IVAS_MAX_NUM_BANDS, SPAR_DIRAC_SPLIT_START_BAND )
		#else
		min( IVAS_MAX_NUM_BANDS, SPAR_DIRAC_SPLIT_START_BAND )
		#endif
		);

		nchan_transport = hMdDec->spar_md_cfg.nchan_transport;

		@@ -525,7 +538,11 @@ ivas_error ivas_spar_md_dec_init(
		ivas_clear_band_coeff_idx( hMdDec->spar_md_prev.band_coeffs_idx_mapped, IVAS_MAX_NUM_BANDS );

		hMdDec->spar_md.dtx_vad = 0;
		#ifdef SBA_HOA_HBR_IMPROV
		hMdDec->spar_md.num_bands = ( hMdDec->spar_hoa_md_flag ) ? IVAS_MAX_NUM_BANDS : min( IVAS_MAX_NUM_BANDS, SPAR_DIRAC_SPLIT_START_BAND );
		#else
		hMdDec->spar_md.num_bands = min( IVAS_MAX_NUM_BANDS, SPAR_DIRAC_SPLIT_START_BAND );
		#endif
		hMdDec->td_decorr_flag = 1;

		set_f( hMdDec->spar_md.en_ratio_slow, 0.0f, IVAS_MAX_NUM_BANDS );
		@@ -683,7 +700,11 @@ void ivas_spar_md_dec_process(
		/* expand DirAC MD to all time slots */
		for ( i_ts = 1; i_ts < MAX_PARAM_SPATIAL_SUBFRAMES; i_ts++ )
		{
		#ifdef SBA_HOA_HBR_IMPROV
		for ( b = 0; b < hMdDec->spar_md.num_bands; b++ )
		#else
		for ( b = 0; b < min( IVAS_MAX_NUM_BANDS, SPAR_DIRAC_SPLIT_START_BAND ); b++ )
		#endif
		{
		for ( j = 0; j < IVAS_SPAR_MAX_CH - 1; j++ )
		{
		@@ -1802,6 +1823,15 @@ static void ivas_decode_arith_bs(
		ndec = hMdDec->spar_md_cfg.num_decorr_per_band[bands_bw * i];

		pred_cell_dims[i].dim1 = ndm + ndec - 1;
		#ifdef SBA_HOA_HBR_IMPROV
		if ( hMdDec->spar_hoa_md_flag )
		{
		if ( i >= SPAR_DIRAC_SPLIT_START_BAND )
		{
		pred_cell_dims[i].dim1 -= ( FOA_CHANNELS - 1 );
		}
		}
		#endif
		pred_cell_dims[i].dim2 = 1;
		drct_cell_dims[i].dim1 = ndec;
		drct_cell_dims[i].dim2 = ndm - 1;
		@@ -1822,6 +1852,23 @@ static void ivas_decode_arith_bs(

		if ( any_diff == 1 )
		{
		#ifdef SBA_HOA_HBR_IMPROV
		if ( hMdDec->spar_hoa_md_flag )
		{
		int16_t j;
		for ( i = 0; i < nB; i++ )
		{
		if ( i >= SPAR_DIRAC_SPLIT_START_BAND )
		{
		for ( j = 0; j < pred_cell_dims[i].dim1; j++ )
		{
		hMdDec->spar_md_prev.band_coeffs_idx_mapped[i].pred_index_re[j] =
		hMdDec->spar_md_prev.band_coeffs_idx_mapped[i].pred_index_re[j + ( FOA_CHANNELS - 1 )];
		}
		}
		}
		}
		#endif
		ivas_copy_band_coeffs_idx_to_arr( hMdDec->spar_md_prev.band_coeffs_idx_mapped, nB, symbol_arr_old_re, pred_cell_dims, PRED_COEFF, planarCP );
		}

		@@ -1830,6 +1877,28 @@ static void ivas_decode_arith_bs(

		ivas_fill_band_coeffs_idx( hMdDec->spar_md.band_coeffs_idx, nB, symbol_arr_re, pred_cell_dims, PRED_COEFF, planarCP );

		#ifdef SBA_HOA_HBR_IMPROV
		if ( hMdDec->spar_hoa_md_flag )
		{
		int16_t j;
		for ( i = 0; i < nB; i++ )
		{
		if ( i >= SPAR_DIRAC_SPLIT_START_BAND )
		{
		for ( j = pred_cell_dims[i].dim1 - 1; j >= 0; j-- )
		{
		hMdDec->spar_md.band_coeffs_idx[i].pred_index_re[j + ( FOA_CHANNELS - 1 )] =
		hMdDec->spar_md.band_coeffs_idx[i].pred_index_re[j];
		}
		for ( j = 0; j < FOA_CHANNELS - 1; j++ )
		{
		hMdDec->spar_md.band_coeffs_idx[i].pred_index_re[j] = 0;
		}
		}
		}
		}
		#endif

		if ( any_diff == 1 )
		{
		ivas_copy_band_coeffs_idx_to_arr( hMdDec->spar_md_prev.band_coeffs_idx_mapped, nB, symbol_arr_old_re, drct_cell_dims, DRCT_COEFF, planarCP );
		@@ -2069,7 +2138,7 @@ static void ivas_decode_huffman_bs(
		for ( i = 0; i < nB; i++ )
		{
		int16_t ndm, ndec;
		int16_t pred_dim, drct_dim, decd_dim;
		int16_t pred_dim, drct_dim, decd_dim, pred_offset;

		ndm = hMdDec->spar_md_cfg.num_dmx_chans_per_band[bands_bw * i];
		ndec = hMdDec->spar_md_cfg.num_decorr_per_band[bands_bw * i];
		@@ -2077,13 +2146,36 @@ static void ivas_decode_huffman_bs(
		pred_dim = ndec + ndm - 1;
		drct_dim = ndec * ( ndm - 1 );
		decd_dim = ndec;
		pred_offset = 0;
		#ifdef SBA_HOA_HBR_IMPROV
		if ( hMdDec->spar_hoa_md_flag )
		{
		if ( i >= SPAR_DIRAC_SPLIT_START_BAND )
		{
		pred_offset = FOA_CHANNELS - 1;
		}
		}
		#endif

		for ( j = 0; j < pred_dim; j++ )
		for ( j = pred_offset; j < pred_dim; j++ )
		{
		ivas_huffman_decode( &hMdDec->huff_coeffs.pred_huff_re[qsi], st0,
		&hMdDec->spar_md.band_coeffs_idx[i].pred_index_re[j] );
		}

		#ifdef SBA_HOA_HBR_IMPROV
		if ( hMdDec->spar_hoa_md_flag )
		{
		if ( i >= SPAR_DIRAC_SPLIT_START_BAND )
		{
		for ( j = 0; j < pred_offset; j++ )
		{
		hMdDec->spar_md.band_coeffs_idx[i].pred_index_re[j] = 0;
		}
		}
		}
		#endif

		for ( j = 0; j < drct_dim; j++ )
		{
		if ( planarCP && !keep_planar[(int16_t) floor( j / ( ndm - 1 ) )] )
		@@ -2761,7 +2853,13 @@ void ivas_spar_to_dirac(
		if ( hMdDec->spar_md_cfg.nchan_transport > 1 )
		{
		ivas_get_spar_md_from_dirac( azi_dirac, ele_dirac, diffuseness, 1, NULL, &hMdDec->spar_md, &hMdDec->spar_md_cfg,
		end_band, num_bands_out, sba_order_internal, dtx_vad, NULL );
		end_band, num_bands_out,
		#ifdef SBA_HOA_HBR_IMPROV
		( hMdDec->spar_hoa_md_flag ) ? 1 : sba_order_internal,
		#else
		sba_order_internal,
		#endif
		dtx_vad, NULL );

		/* temporarily copy frame-wise prediction coefficients in DirAC bands*/
		for ( pred_idx = 0; pred_idx < FOA_CHANNELS - 1; pred_idx++ )
		@@ -2774,7 +2872,13 @@ void ivas_spar_to_dirac(
		}

		ivas_get_spar_md_from_dirac( azi_dirac, ele_dirac, diffuseness, MAX_PARAM_SPATIAL_SUBFRAMES, NULL, &hMdDec->spar_md, &hMdDec->spar_md_cfg,
		end_band, num_bands_out, sba_order_internal, dtx_vad, NULL );
		end_band, num_bands_out,
		#ifdef SBA_HOA_HBR_IMPROV
		( hMdDec->spar_hoa_md_flag ) ? 1 : sba_order_internal,
		#else
		sba_order_internal,
		#endif
		dtx_vad, NULL );

		/* expand DirAC TC 20ms MD for residual channels to all subframes*/
		for ( block = 0; block < MAX_PARAM_SPATIAL_SUBFRAMES; block++ )