Adding in mono detection and processing to stop leakage (ce428ca7) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/ivas_prot.h

+18 −0

Original line number	Diff line number	Diff line
		@@ -3375,6 +3375,11 @@ void ivas_dirac_param_est_enc(
		const IVAS_FORMAT ivas_format ,
		const int16_t hodirac_flag,
		const int16_t nchan_fb_in
		#ifdef FIX_527_SBA_MONO_INPUT
		,
		int16_t *mono_frame_count,
		int16_t *dirac_mono_flag
		#endif
		);


		@@ -4327,6 +4332,9 @@ ivas_error ivas_spar_md_enc_process(
		const int16_t nchan_inp,
		const int16_t sba_order, /* i : Ambisonic (SBA) order */
		float prior_mixer[IVAS_MAX_FB_MIXER_OUT_CH][IVAS_MAX_SPAR_FB_MIXER_IN_CH] / i : prior mixer_matrix */
		#ifdef FIX_527_SBA_MONO_INPUT
		,const int16_t dirac_mono_flag
		#endif
		);

		void ivas_compute_spar_params(
		@@ -4466,6 +4474,11 @@ void ivas_spar_update_md_hist(
		ivas_spar_md_dec_state_t hMdDec / i/o: SPAR MD decoder handle */
		);

		int16_t ivas_spar_chk_zero_coefs(
		Decoder_Struct st_ivas, / i/o: IVAS decoder handle */
		const int16_t sba_order /* i : Ambisonic (SBA) order */
		);

		void ivas_spar_smooth_md_dtx(
		ivas_spar_md_dec_state_t hMdDec, / i/o: SPAR MD decoder handle */
		const int16_t num_bands_out, /* i : number of output bands */
		@@ -5345,6 +5358,11 @@ void computeReferencePower_enc(
		const IVAS_FORMAT ivas_format, /* i : ivas_format */
		int16_t ref_power_w, /* i : use 0 if hodirac is enabled */
		const int16_t nchan_ana /* i : number of analysis channels */
		#ifdef FIX_527_SBA_MONO_INPUT
		,
		int16_t *mono_frame_count,
		int16_t *dirac_mono_flag
		#endif
		);

		ivas_error ivas_mono_dmx_renderer_open(

lib_com/ivas_stat_com.h

+3 −0

Original line number	Diff line number	Diff line
		@@ -540,6 +540,9 @@ typedef struct ivas_masa_qmetadata_frame_struct
		int16_t metadata_max_bits; /* maximum allowed number of bits for metadata per frame */
		uint8_t useLowerRes;
		uint8_t useLowerBandRes;
		#ifdef FIX_527_SBA_MONO_INPUT
		int16_t dirac_mono_flag;
		#endif
		IVAS_SURROUND_COHERENCE_BAND_DATA *surcoh_band_data;

		/* Additional helper values to include all data required for writing to output file */

lib_com/options.h

+2 −1

Original line number	Diff line number	Diff line
		@@ -210,8 +210,9 @@
		#define FIX_626_VARIABLE_TYPE_MDCT_CONC /* FhG: trivial fix to fix USAN error */
		#define FIX_616_DIV_ZERO_MCT /FhG : Fix UBSAN division by zero error of issue 616/

		/* ################## End BE DEVELOPMENT switches ######################### */
		#define FIX_527_SBA_MONO_INPUT /Dlb : Fix for mono content in a HOA input format /

		/* ################## End BE DEVELOPMENT switches ######################### */

		/* #################### Start NON-BE CR switches ########################## */
		/* any switch which is non-be wrt operation points tested in selection */

lib_dec/ivas_spar_md_dec.c

+120 −0

Original line number	Diff line number	Diff line
		@@ -725,6 +725,63 @@ static ivas_error ivas_spar_set_dec_config(
		return IVAS_ERR_OK;
		}

		/-----------------------------------------------------------------------------------------
		* Function ivas_dec_mono_sba_handling()
		*
		*
		-----------------------------------------------------------------------------------------/

		static void ivas_dec_mono_sba_handling(
		Decoder_Struct st_ivas, / i/o: IVAS decoder handle */
		const int16_t sba_order /* i : Ambisonic (SBA) order */
		)
		{
		int16_t mono_flag, b, block;

		mono_flag = 1;

		for ( b = 0; b < st_ivas->hQMetaData->q_direction[0].cfg.nbands; b++ )
		{
		for ( block = 0; block < MAX_PARAM_SPATIAL_SUBFRAMES; ++block )
		{
		float azimuth = st_ivas->hQMetaData->q_direction[0].band_data[b].azimuth[block];
		float elevation = st_ivas->hQMetaData->q_direction[0].band_data[b].azimuth[block];
		float energy_ratio = st_ivas->hQMetaData->q_direction[0].band_data[0].energy_ratio[block];
		if (
		( azimuth != 0.0f ) \|\|
		( elevation != 0.0f ) \|\|
		( energy_ratio > 0.15f ) ) /* 0.15f is just above the lowest quantised value. */
		{
		mono_flag = 0;
		}
		}
		}
		/* Combine the SPAR prediction coefs flag with the azimuth, elevation and energy ratio flag.*/
		mono_flag = mono_flag && ivas_spar_chk_zero_coefs( st_ivas, sba_order );

		if ( mono_flag )
		{
		/* Set Energy Ratio values to be zero */
		for ( b = 0; b < st_ivas->hQMetaData->q_direction[0].cfg.nbands; b++ )
		{
		set_zero( st_ivas->hQMetaData->q_direction[0].band_data[b].energy_ratio, MAX_PARAM_SPATIAL_SUBFRAMES );
		}
		if ( st_ivas->hDirAC != NULL )
		{
		for ( block = 0; block < st_ivas->hSpatParamRendCom->dirac_md_buffer_length; ++block )
		{
		/* Set directional Energy Ratio values to be zero */
		set_zero( st_ivas->hSpatParamRendCom->energy_ratio1[block], st_ivas->hSpatParamRendCom->num_freq_bands );
		if ( st_ivas->hQMetaData->no_directions == 2 )
		{
		set_zero( st_ivas->hSpatParamRendCom->energy_ratio2[block], st_ivas->hSpatParamRendCom->num_freq_bands );
		}
		/* Set Diffuseness values to be 1.0 */
		set_f( st_ivas->hSpatParamRendCom->diffuseness_vector[block], 1.0f, st_ivas->hSpatParamRendCom->num_freq_bands );
		}
		}
		}
		}

		/-----------------------------------------------------------------------------------------
		* Function ivas_spar_md_dec_process()
		@@ -764,6 +821,10 @@ void ivas_spar_md_dec_process(
		#endif
		st_ivas->hQMetaData->sba_inactive_mode, st_ivas->last_active_ivas_total_brate );

		#ifdef FIX_527_SBA_MONO_INPUT
		ivas_dec_mono_sba_handling( st_ivas, sba_order );
		#endif

		#if 0
		{
		char f_name[100];
		@@ -1090,7 +1151,66 @@ void ivas_spar_md_dec_process(

		return;
		}
		/-----------------------------------------------------------------------------------------
		* Function ivas_spar_chk_zero_coefs()
		*
		* Check for zeroed SPAR coefficients
		-----------------------------------------------------------------------------------------/
		int16_t ivas_spar_chk_zero_coefs(
		Decoder_Struct st_ivas, / i/o: IVAS decoder handle */
		const int16_t sba_order /* i : Ambisonic (SBA) order */
		)
		{
		int16_t j, k, b, i_ts;
		ivas_spar_md_dec_state_t *hMdDec;
		int16_t num_md_sub_frames;
		int16_t mono = 1;

		hMdDec = st_ivas->hSpar->hMdDec;

		num_md_sub_frames = ivas_get_spar_dec_md_num_subframes( sba_order, st_ivas->hDecoderConfig->ivas_total_brate
		#ifdef VLBR_20MS_MD
		,
		st_ivas->last_active_ivas_total_brate
		#endif
		);

		int16_t ndec = hMdDec->spar_md_cfg.num_decorr_per_band[0];
		int16_t ndm = hMdDec->spar_md_cfg.num_dmx_chans_per_band[0];

		for ( i_ts = 0; i_ts < num_md_sub_frames; i_ts++ )
		{
		for ( b = 0; b < min( hMdDec->spar_md.num_bands, SPAR_DIRAC_SPLIT_START_BAND ); b++ )
		{
		for ( j = 0; j < ndm + ndec - 1; j++ )
		{
		// printf("%e, ", hMdDec->spar_md.band_coeffs[b + i_ts * IVAS_MAX_NUM_BANDS].pred_re[j]);
		if ( hMdDec->spar_md.band_coeffs[b + i_ts * IVAS_MAX_NUM_BANDS].pred_re[j] != 0.0f )
		{
		mono = 0;
		}
		}
		for ( j = 0; j < ndec; j++ )
		{
		for ( k = 0; k < ndm - 1; k++ )
		{
		if ( hMdDec->spar_md.band_coeffs[b + i_ts * IVAS_MAX_NUM_BANDS].C_re[j][k] != 0.0f )
		{
		mono = 0;
		}
		}
		}
		for ( j = 0; j < ndec; j++ )
		{
		if ( hMdDec->spar_md.band_coeffs[b + i_ts * IVAS_MAX_NUM_BANDS].P_re[j] != 0.0f )
		{
		mono = 0;
		}
		}
		}
		}
		return mono;
		}

		/-----------------------------------------------------------------------------------------
		* Function ivas_spar_smooth_md_dtx()

lib_enc/ivas_dirac_enc.c

+171 −5

Original line number	Diff line number	Diff line
		@@ -159,6 +159,7 @@ ivas_error ivas_dirac_enc_open(
		}

		hDirAC->index_buffer_intensity = 0;
		hDirAC->mono_frame_count = 0;

		st_ivas->hDirAC = hDirAC;
		st_ivas->hSpar->enc_param_start_band = st_ivas->hDirAC->hConfig->enc_param_start_band;
		@@ -299,7 +300,13 @@ void ivas_dirac_enc(
		int16_t i, j, b, i_ts;
		push_wmops( "ivas_dirac_enc" );

		ivas_dirac_param_est_enc( hDirAC, hQMetaData->q_direction, hQMetaData->useLowerRes, data_f, ppIn_FR_real, ppIn_FR_imag, input_frame, ivas_format, hodirac_flag, hodirac_flag ? HOA2_CHANNELS : FOA_CHANNELS );
		ivas_dirac_param_est_enc( hDirAC, hQMetaData->q_direction, hQMetaData->useLowerRes, data_f, ppIn_FR_real, ppIn_FR_imag, input_frame, ivas_format, hodirac_flag, hodirac_flag ? HOA2_CHANNELS : FOA_CHANNELS
		#ifdef FIX_527_SBA_MONO_INPUT
		,
		&( hDirAC->mono_frame_count ),
		&( hQMetaData->dirac_mono_flag )
		#endif
		);

		if ( hQMetaData->q_direction->cfg.nbands > 0 )
		{
		@@ -310,6 +317,22 @@ void ivas_dirac_enc(
		/* WB 4TC mode bit : disable for now*/
		push_next_indice( hMetaData, 0, 1 );

		#ifdef FIX_527_SBA_MONO_INPUT
		/* Set Energy Ratio to 0.0 if the mono flag is set, before the metadata is encoded */
		if ( hQMetaData->dirac_mono_flag )
		{
		for ( b = hQMetaData->q_direction->cfg.start_band; b < hQMetaData->q_direction->cfg.nbands; b++ )
		{
		for ( i_ts = 0; i_ts < ( ( dtx_vad == 1 ) ? hQMetaData->q_direction[0].cfg.nblocks : 1 ); i_ts++ )
		{
		hQMetaData->q_direction[0].band_data[b].energy_ratio[i_ts] = 0.0f;
		hQMetaData->q_direction[0].band_data[b].azimuth[i_ts] = 0.0f;
		hQMetaData->q_direction[0].band_data[b].elevation[i_ts] = 0.0f;
		}
		}
		}
		#endif

		ivas_qmetadata_enc_encode( hMetaData, hQMetaData, hodirac_flag );
		}
		else
		@@ -388,6 +411,125 @@ void ivas_dirac_enc(
		return;
		}

		/*-------------------------------------------------------------------------
		* ivas_dirac_get_mono_flag()
		*
		*
		-------------------------------------------------------------------------/

		static int16_t ivas_dirac_get_mono_flag(
		const int16_t band_grouping, / i : Band grouping for estimation */
		float Cldfb_RealBuffer[DIRAC_MAX_ANA_CHANS][DIRAC_NO_FB_BANDS_MAX], /* i : Real part of input signal */
		float Cldfb_ImagBuffer[DIRAC_MAX_ANA_CHANS][DIRAC_NO_FB_BANDS_MAX], /* i : Imag part of input signal */
		const int16_t nchan_ana, /* i : number of analysis channels */
		int16_t mono_frame_count ) / i/o : current number of mono frames count */
		{
		int16_t brange[2];
		int16_t i, j, ch_idx;
		float other_ch_band_power;
		float W_band_power;
		int16_t any_mc_band = 0;
		int16_t any_mono_band = 0;
		int16_t local_mono_flag = 0;
		float W_band_power_norm;
		float threshold = 0;

		static const float THRESH_SILENCE = 3e4f;
		static const float NORM_FACTOR = 1e13f;
		static const float ONE_ON_NORM_FACTOR = 1 / 1e13f;
		static const float MAX_THRESH = 1e6f;
		static const float MIN_THRESH = 1e2f;
		static const float MONO_FRAME_THRESH = ( 15 ); /* 30ms */

		/* Banded Power Calculations */
		for ( i = 0; i < DIRAC_MAX_NBANDS; i++ )
		{
		W_band_power = 0;
		other_ch_band_power = 0;

		brange[0] = band_grouping[i];
		brange[1] = band_grouping[i + 1];

		/* Loop over the W channel bins to calculate the power in the band */
		for ( j = brange[0]; j < brange[1]; j++ )
		{
		W_band_power += ( Cldfb_RealBuffer[0][j] * Cldfb_RealBuffer[0][j] ) + ( Cldfb_ImagBuffer[0][j] * Cldfb_ImagBuffer[0][j] );
		}

		/* Loop over the other channels and bins to calculate the power in the band */
		for ( ch_idx = 1; ch_idx < nchan_ana; ch_idx++ )
		{
		/* abs()^2 */
		for ( j = brange[0]; j < brange[1]; j++ )
		{
		other_ch_band_power += ( Cldfb_RealBuffer[ch_idx][j] * Cldfb_RealBuffer[ch_idx][j] ) + ( Cldfb_ImagBuffer[ch_idx][j] * Cldfb_ImagBuffer[ch_idx][j] );
		}
		}
		if ( other_ch_band_power < EPSILON )
		{
		if ( W_band_power > THRESH_SILENCE )
		{
		any_mono_band = 1;
		}
		}
		else
		{
		if ( ( W_band_power > THRESH_SILENCE ) \|\| ( other_ch_band_power > THRESH_SILENCE ) )
		{
		W_band_power_norm = min( W_band_power, NORM_FACTOR ) * ONE_ON_NORM_FACTOR;
		threshold = max( W_band_power_norm * MAX_THRESH, MIN_THRESH );
		if ( W_band_power / other_ch_band_power > threshold )
		{
		any_mono_band = 1;
		}
		else
		{
		any_mc_band = 1;
		}
		}
		}
		}
		/* If any band contains multi-channel content it's not mono */
		if ( any_mc_band )
		{
		local_mono_flag = 0;
		}
		else
		{
		/* If any band contains mono content the frame is mono. */
		if ( any_mono_band )
		{
		local_mono_flag = 1;
		}
		}
		/* Hysteresis - only after MONO_FRAME_THRESH frames on mono will the actual mono flag be set */
		if ( local_mono_flag )
		{
		if ( *mono_frame_count < MONO_FRAME_THRESH )
		{
		( *mono_frame_count )++;
		}
		}
		else
		{
		/* Instantaneously disable actual mono flag if multi-channel content is observed */
		if ( any_mc_band )
		{
		*mono_frame_count = 0;
		}
		}

		/* Final check if there has been mono for MONO_FRAME_THRESH number of frames than the content is declared mono */
		if ( *mono_frame_count == MONO_FRAME_THRESH )
		{
		return 1;
		}
		else
		{
		return 0;
		}
		}

		/*-------------------------------------------------------------------------
		* computeReferencePower_enc()
		*
		@@ -404,6 +546,11 @@ void computeReferencePower_enc(
		const IVAS_FORMAT ivas_format, /* i : ivas_format */
		int16_t ref_power_w, /* i : use 0 if hodirac is enabled */
		const int16_t nchan_ana /* i : number of analysis channels */
		#ifdef FIX_527_SBA_MONO_INPUT
		,
		int16_t mono_frame_count, / i/o: Mono Frame Count */
		int16_t dirac_mono_flag / i/o: Mono Flag */
		#endif
		)
		{
		int16_t brange[2];
		@@ -411,10 +558,18 @@ void computeReferencePower_enc(

		float reference_power_W[DIRAC_MAX_NBANDS];

		#ifdef FIX_527_SBA_MONO_INPUT
		if ( dirac_mono_flag != NULL )
		{
		*dirac_mono_flag = ivas_dirac_get_mono_flag( band_grouping, Cldfb_RealBuffer, Cldfb_ImagBuffer, nchan_ana, mono_frame_count );
		}
		#endif

		for ( i = 0; i < num_freq_bands; i++ )
		{
		brange[0] = band_grouping[i + enc_param_start_band];
		brange[1] = band_grouping[i + enc_param_start_band + 1];

		reference_power[i] = 0;

		reference_power_W[i] = 0;
		@@ -446,7 +601,6 @@ void computeReferencePower_enc(
		return;
		}


		/*-------------------------------------------------------------------------
		* ivas_dirac_param_est_enc()
		*
		@@ -463,7 +617,13 @@ void ivas_dirac_param_est_enc(
		const int16_t input_frame,
		const IVAS_FORMAT ivas_format,
		const int16_t hodirac_flag,
		const int16_t nchan_fb_in )
		const int16_t nchan_fb_in
		#ifdef FIX_527_SBA_MONO_INPUT
		,
		int16_t *mono_frame_count,
		int16_t *dirac_mono_flag
		#endif
		)
		{
		int16_t i, d, ts, index, l_ts, num_freq_bands;
		int16_t band_m_idx, block_m_idx;
		@@ -571,7 +731,13 @@ void ivas_dirac_param_est_enc(
		num_freq_bands,
		ivas_format,
		hodirac_flag ? 0 : 1,
		FOA_CHANNELS );
		FOA_CHANNELS
		#ifdef FIX_527_SBA_MONO_INPUT
		,
		mono_frame_count,
		dirac_mono_flag
		#endif
		);

		computeIntensityVector_enc(
		hDirAC,