implement new prediction formula with sector DoAs (e1049aef) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_prot.h

+9 −0

Original line number	Diff line number	Diff line
		@@ -4185,6 +4185,11 @@ void ivas_get_spar_md_from_dirac(
		float azi_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float ele_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float diffuseness[IVAS_MAX_NUM_BANDS],
		#ifdef HODIRAC
		float azi_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float ele_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float ratio[IVAS_MAX_NUM_BANDS],
		#endif
		const int16_t n_ts,
		float ***mixer_mat,
		ivas_spar_md_t *hSpar_md,
		@@ -4194,6 +4199,10 @@ void ivas_get_spar_md_from_dirac(
		const int16_t order,
		const int16_t dtx_vad,
		float Wscale_d[IVAS_MAX_NUM_BANDS]
		#ifdef HODIRAC
		,
		int16_t hodirac
		#endif
		);

		ivas_error ivas_spar_md_dec_open(

lib_com/ivas_spar_com.c

+117 −3

Original line number	Diff line number	Diff line
		@@ -1640,6 +1640,11 @@ void ivas_get_spar_md_from_dirac(
		float azi_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float ele_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float diffuseness[IVAS_MAX_NUM_BANDS],
		#ifdef HODIRAC
		float azi_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float ele_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES],
		float ratio[IVAS_MAX_NUM_BANDS],
		#endif
		const int16_t n_ts,
		float ***mixer_mat,
		ivas_spar_md_t *hSpar_md,
		@@ -1648,12 +1653,21 @@ void ivas_get_spar_md_from_dirac(
		const int16_t end_band,
		const int16_t order,
		const int16_t dtx_vad,
		float Wscale_d[IVAS_MAX_NUM_BANDS] )
		float Wscale_d[IVAS_MAX_NUM_BANDS]
		#ifdef HODIRAC
		,
		int16_t hodirac
		#endif
		)
		{
		int16_t num_ch, band, i, j;
		int16_t block, ch;
		float response_avg[MAX_OUTPUT_CHANNELS];
		float response[MAX_PARAM_SPATIAL_SUBFRAMES][MAX_OUTPUT_CHANNELS];
		#ifdef HODIRAC
		float response_avg2[MAX_OUTPUT_CHANNELS];
		float response2[MAX_PARAM_SPATIAL_SUBFRAMES][MAX_OUTPUT_CHANNELS];
		#endif
		float cov_real_dirac[IVAS_SPAR_MAX_CH][IVAS_SPAR_MAX_CH][IVAS_MAX_NUM_BANDS];
		float *pCov_real[IVAS_SPAR_MAX_CH][IVAS_SPAR_MAX_CH];
		float dm_fv_re[IVAS_SPAR_MAX_CH - 1][IVAS_MAX_NUM_BANDS];
		@@ -1662,6 +1676,10 @@ void ivas_get_spar_md_from_dirac(
		float **ppMixer_mat[IVAS_MAX_FB_MIXER_OUT_CH];
		float *pMixer_mat[IVAS_MAX_FB_MIXER_OUT_CH][IVAS_MAX_SPAR_FB_MIXER_IN_CH];
		float en_ratio_fac, diff_norm_order1, diff_norm_order2, diff_norm_order3;
		#ifdef HODIRAC
		float w1 = 0;
		float w2 = 0;
		#endif
		int16_t ndm, foa_ch, hoa2_ch;
		float P_dir_fact[IVAS_SPAR_MAX_CH - 1];
		const int16_t *remix_order;
		@@ -1743,15 +1761,34 @@ void ivas_get_spar_md_from_dirac(

		/SPAR from DirAC/
		set_f( response_avg, 0.0f, MAX_OUTPUT_CHANNELS );
		#ifdef HODIRAC
		if ( hodirac )
		{
		set_f( response_avg2, 0.0f, MAX_OUTPUT_CHANNELS );
		}
		#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 );
		#ifdef HODIRAC
		if ( hodirac )
		{
		ivas_dirac_dec_get_response( (int16_t) azi_dirac2[band][i_ts], (int16_t) ele_dirac2[band][i_ts], response_avg2, order );
		}
		#endif
		}
		else
		{
		for ( block = 0; block < MAX_PARAM_SPATIAL_SUBFRAMES; block++ )
		{
		ivas_dirac_dec_get_response( (int16_t) azi_dirac[band][block], (int16_t) ele_dirac[band][block], &( response[block][0] ), order );
		#ifdef HODIRAC
		if ( hodirac )
		{
		ivas_dirac_dec_get_response( (int16_t) azi_dirac2[band][block], (int16_t) ele_dirac2[band][block], &( response2[block][0] ), order );
		}
		#endif
		}

		/* average responses in all subframes*/
		@@ -1806,6 +1843,56 @@ void ivas_get_spar_md_from_dirac(
		{
		response_avg[ch] /= norm;
		}

		#ifdef HODIRAC
		if ( hodirac )
		{

		for ( ch = 0; ch < num_ch_order; ch++ )
		{
		for ( block = 0; block < MAX_PARAM_SPATIAL_SUBFRAMES; block++ )
		{
		response_avg2[ch] += response2[block][ch];
		}
		response_avg2[ch] /= MAX_PARAM_SPATIAL_SUBFRAMES;
		}

		norm = 0.0f;
		for ( ch = 1; ch < foa_ch; ch++ )
		{
		norm += response_avg2[ch] * response_avg2[ch];
		}
		norm = max( EPSILON, sqrtf( norm ) );
		for ( ch = 1; ch < foa_ch; ch++ )
		{
		response_avg2[ch] /= norm;
		}

		/normalize 2nd order/
		norm = 0.0f;
		for ( ch = foa_ch; ch < min( hoa2_ch_order, num_ch_order ); ch++ )
		{
		norm += response_avg2[ch] * response_avg2[ch];
		}
		norm = max( EPSILON, sqrtf( norm ) );
		for ( ch = foa_ch; ch < min( hoa2_ch_order, num_ch_order ); ch++ )
		{
		response_avg2[ch] /= norm;
		}

		/normalize 3rd order/
		norm = 0.0f;
		for ( ch = hoa2_ch_order; ch < num_ch_order; ch++ )
		{
		norm += response_avg2[ch] * response_avg2[ch];
		}
		norm = max( EPSILON, sqrtf( norm ) );
		for ( ch = hoa2_ch_order; ch < num_ch_order; ch++ )
		{
		response_avg2[ch] /= norm;
		}
		}
		#endif
		}
		}

		@@ -1813,7 +1900,23 @@ void ivas_get_spar_md_from_dirac(
		{
		response_avg[i] = response_avg[HOA_keep_ind[i]];
		}
		#ifdef HODIRAC
		if ( hodirac )
		{
		for ( i = FOA_CHANNELS + 1; i < num_ch; i++ )
		{
		response_avg2[i] = response_avg2[HOA_keep_ind[i]];
		}
		}
		#endif
		en_ratio_fac = ( 1.0f - diffuseness[band] );
		#ifdef HODIRAC
		if ( hodirac )
		{
		w1 = ratio[band] * ratio[band];
		w2 = ( 1.0f - ratio[band] ) * ( 1.0f - ratio[band] );
		}
		#endif

		for ( i = 0; i < num_ch; i++ )
		{
		@@ -1828,8 +1931,10 @@ void ivas_get_spar_md_from_dirac(
		else
		{
		cov_real_dirac[i][j][band] = en_ratio_fac * response_avg[i] * response_avg[j];

		if ( hSpar_md_cfg->nchan_transport <= 2 )
		{

		cov_real_dirac[i][j][band] *= en_ratio_fac;
		if ( ( i >= ndm ) && ( dtx_vad == 1 ) )
		{
		@@ -1870,11 +1975,20 @@ void ivas_get_spar_md_from_dirac(
		}
		else
		{
		#ifdef HODIRAC
		if ( hodirac )
		{
		cov_real_dirac[i][j][band] = en_ratio_fac * ( w1 * response_avg[i] * response_avg[j] + w2 * response_avg2[i] * response_avg2[j] );
		}
		else
		#endif
		{
		cov_real_dirac[i][j][band] = en_ratio_fac * response_avg[i] * response_avg[j];
		}
		}
		}
		}
		}

		for ( i = 0; i < num_ch; i++ )
		{
		@@ -1884,7 +1998,7 @@ void ivas_get_spar_md_from_dirac(
		}
		}

		#ifdef DEBUG_SPAR_WRITE_OUT_COV
		#ifdef DEBUG_SBA_MD_DUMP
		{
		static FILE *fid = 0;
		int16_t k = 0;

lib_com/options.h

+2 −2

Original line number	Diff line number	Diff line
		@@ -151,11 +151,11 @@
		#define HODIRAC /* FhG: Sector-based HO-DirAC method for SBA at high bitrates */
		#define HODIRAC_CRASH_FIX

		/#define HODIRAC_DEBUG/
		#define HODIRAC_DEBUG

		#ifdef HODIRAC_DEBUG
		//#define HODIRAC_CHECK_VALUE_RANGE
		//#define HODIRAC_WRITE_PARAMS
		#define HODIRAC_WRITE_PARAMS
		///#define HODIRAC_READ_PARAMS
		#endif

lib_dec/ivas_spar_md_dec.c

+65 −3

Original line number	Diff line number	Diff line
		@@ -1389,7 +1389,7 @@ void ivas_spar_dec_gen_umx_mat(
		}
		}

		#if DEBUG_SBA_MD_DUMP
		#ifdef DEBUG_SBA_MD_DUMP
		{
		static FILE *f_mat = 0;

		@@ -2542,6 +2542,11 @@ void ivas_spar_to_dirac(
		int16_t sba_order_internal;
		float azi_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		float ele_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		#ifdef HODIRAC
		float azi_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		float ele_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		float ratio[IVAS_MAX_NUM_BANDS];
		#endif
		int16_t azi[IVAS_MAX_NUM_BANDS];
		int16_t ele[IVAS_MAX_NUM_BANDS];
		float dvx[IVAS_MAX_NUM_BANDS], dvy[IVAS_MAX_NUM_BANDS], dvz[IVAS_MAX_NUM_BANDS];
		@@ -2739,15 +2744,50 @@ void ivas_spar_to_dirac(
		}
		azi_dirac[band][block] = st_ivas->hQMetaData->q_direction->band_data[dirac_band_idx].azimuth[block];
		ele_dirac[band][block] = st_ivas->hQMetaData->q_direction->band_data[dirac_band_idx].elevation[block];
		#ifdef HODIRAC
		if ( st_ivas->sba_analysis_order > 1 )
		{
		azi_dirac2[band][block] = st_ivas->hQMetaData->q_direction[1].band_data[dirac_band_idx].azimuth[block];
		ele_dirac2[band][block] = st_ivas->hQMetaData->q_direction[1].band_data[dirac_band_idx].elevation[block];
		}
		#endif
		}

		diffuseness[band] = 1.0f - st_ivas->hQMetaData->q_direction->band_data[dirac_band_idx].energy_ratio[0];
		#ifdef HODIRAC
		if ( st_ivas->sba_analysis_order > 1 )
		{
		ratio[band] = st_ivas->hQMetaData->q_direction[1].band_data[dirac_band_idx].energy_ratio[0];
		}
		#endif
		}

		/* DirAC MD averaged over 4 subframes and converted to SPAR format similar to encoder processing */
		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, ( hMdDec->spar_hoa_md_flag ) ? 1 : sba_order_internal, dtx_vad, NULL );
		ivas_get_spar_md_from_dirac(
		azi_dirac,
		ele_dirac,
		diffuseness,
		#ifdef HODIRAC
		azi_dirac2,
		ele_dirac2,
		ratio,
		#endif
		1,
		NULL,
		&hMdDec->spar_md,
		&hMdDec->spar_md_cfg,
		end_band,
		num_bands_out,
		( hMdDec->spar_hoa_md_flag ) ? 1 : sba_order_internal,
		dtx_vad,
		NULL
		#ifdef HODIRAC
		,
		st_ivas->sba_analysis_order > 1
		#endif
		);

		/* temporarily copy frame-wise prediction coefficients in DirAC bands*/
		for ( pred_idx = 0; pred_idx < FOA_CHANNELS - 1; pred_idx++ )
		@@ -2759,7 +2799,29 @@ 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, ( hMdDec->spar_hoa_md_flag ) ? 1 : sba_order_internal, dtx_vad, NULL );
		ivas_get_spar_md_from_dirac(
		azi_dirac,
		ele_dirac,
		diffuseness,
		#ifdef HODIRAC
		azi_dirac2,
		ele_dirac2,
		ratio,
		#endif
		MAX_PARAM_SPATIAL_SUBFRAMES,
		NULL,
		&hMdDec->spar_md,
		&hMdDec->spar_md_cfg,
		end_band,
		num_bands_out,
		( hMdDec->spar_hoa_md_flag ) ? 1 : sba_order_internal,
		dtx_vad,
		NULL
		#ifdef HODIRAC
		,
		st_ivas->sba_analysis_order > 1
		#endif
		);

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

lib_enc/ivas_spar_encoder.c

+49 −3

Original line number	Diff line number	Diff line
		@@ -721,6 +721,11 @@ static ivas_error ivas_spar_enc_process(
		{
		float azi_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		float ele_dirac[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		#ifdef HODIRAC
		float azi_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		float ele_dirac2[IVAS_MAX_NUM_BANDS][MAX_PARAM_SPATIAL_SUBFRAMES];
		float ratio[IVAS_MAX_NUM_BANDS];
		#endif
		float diffuseness[IVAS_MAX_NUM_BANDS];
		float Wscale_d[IVAS_MAX_NUM_BANDS];
		int16_t d_start_band, d_end_band;
		@@ -734,10 +739,31 @@ static ivas_error ivas_spar_enc_process(
		dirac_band_idx = hSpar->dirac_to_spar_md_bands[b] - d_start_band;
		for ( i_ts = 0; i_ts < hQMetaData->q_direction->cfg.nblocks; i_ts++ )
		{
		azi_dirac[b][i_ts] = hQMetaData->q_direction->band_data[dirac_band_idx].azimuth[i_ts];
		ele_dirac[b][i_ts] = hQMetaData->q_direction->band_data[dirac_band_idx].elevation[i_ts];
		azi_dirac[b][i_ts] = hQMetaData->q_direction[0].band_data[dirac_band_idx].azimuth[i_ts];
		ele_dirac[b][i_ts] = hQMetaData->q_direction[0].band_data[dirac_band_idx].elevation[i_ts];
		#ifdef HODIRAC
		if ( st_ivas->sba_analysis_order > 1 )
		{
		azi_dirac2[b][i_ts] = hQMetaData->q_direction[1].band_data[dirac_band_idx].azimuth[i_ts] + hQMetaData->q_direction[0].band_data[dirac_band_idx].azimuth[i_ts] - 180.f;
		if ( azi_dirac2[b][i_ts] < -180.f )
		{
		azi_dirac2[b][i_ts] += 360.f;
		}
		if ( azi_dirac2[b][i_ts] >= 180.f )
		{
		azi_dirac2[b][i_ts] -= 360.f;
		}
		ele_dirac2[b][i_ts] = hQMetaData->q_direction[1].band_data[dirac_band_idx].elevation[i_ts];
		}
		#endif
		}
		diffuseness[b] = 1.0f - hQMetaData->q_direction->band_data[dirac_band_idx].energy_ratio[0];
		#ifdef HODIRAC
		if ( st_ivas->sba_analysis_order > 1 )
		{
		ratio[b] = hQMetaData->q_direction[1].band_data[dirac_band_idx].energy_ratio[0];
		}
		#endif
		}

		if ( d_start_band >= 6 && dtx_vad == 1 )
		@@ -757,7 +783,27 @@ static ivas_error ivas_spar_enc_process(
		Wscale_d[b] = min( 2.0f, max( Wscale_d[b], 1.0f ) );
		}

		ivas_get_spar_md_from_dirac( azi_dirac, ele_dirac, diffuseness, 1, hSpar->hMdEnc->mixer_mat, &hSpar->hMdEnc->spar_md, &hSpar->hMdEnc->spar_md_cfg, d_start_band, d_end_band, ( hSpar->hMdEnc->spar_hoa_md_flag ) ? 1 : sba_order, dtx_vad, Wscale_d );
		ivas_get_spar_md_from_dirac(
		azi_dirac,
		ele_dirac,
		diffuseness,
		#ifdef HODIRAC
		azi_dirac2,
		ele_dirac2,
		ratio,
		#endif
		1,
		hSpar->hMdEnc->mixer_mat,
		&hSpar->hMdEnc->spar_md,
		&hSpar->hMdEnc->spar_md_cfg,
		d_start_band,
		d_end_band,
		( hSpar->hMdEnc->spar_hoa_md_flag ) ? 1 : sba_order, dtx_vad, Wscale_d
		#ifdef HODIRAC
		,
		st_ivas->sba_analysis_order > 1
		#endif
		);
		}

		if ( hSpar->hMdEnc->spar_hoa_md_flag )