Merge branch 'main' into 573-fix_dirac_ls_synthesis_config-leftover-in-the-code (0ecb70c2) · Commits · IVAS Codec Public Collaboration / IVAS Codec

.gitlab-ci.yml

+40 −0

Original line number	Diff line number	Diff line
		@@ -10,6 +10,7 @@ variables:
		OUT_FORMATS_BINAURAL: "BINAURAL BINAURAL_ROOM_IR BINAURAL_ROOM_REVERB"
		EXIT_CODE_NON_BE: 123
		EXIT_CODE_FAIL: 1
		PROCESSING_SCRIPTS_BIN_DIR: "/test-bin"

		default:
		interruptible: true # Make all jobs by default interruptible
		@@ -692,6 +693,45 @@ check-first-frame-is-sid:
		expose_as: "logs-sidstart"
		expire_in: "5 days"

		selection-test-processing:
		extends:
		- .test-job-linux-needs-testv-dir
		- .rules-merge-request
		tags:
		- processing-scripts-linux
		stage: test
		needs: ["build-codec-linux-make"]
		script:
		# get processing script code
		- git clone https://forge.3gpp.org/rep/ivas-codec-pc/ivas-processing-scripts.git --single-branch -b main

		- mkdir -p ivas-processing-scripts/tests/data/testv
		- cp -r scripts/testv/* ivas-processing-scripts/tests/data/testv/

		# copy binaries into local bin dir, those should take precendence over PATH
		- cp $PROCESSING_SCRIPTS_BIN_DIR/* ivas-processing-scripts/ivas_processing_scripts/bin/
		# for testing with native binaries
		- rm ivas-processing-scripts/ivas_processing_scripts/bin/IVAS*.exe
		- rm ivas-processing-scripts/ivas_processing_scripts/bin/EVS*.exe

		# build codec and put into bin dir
		- make -j
		- cp ./IVAS_* ivas-processing-scripts/ivas_processing_scripts/bin/

		# patch the use_windows_codec_binaries key (weird folding is needed so colons are accepted)
		- >
		sed -i "s/use_windows_codec_binaries: true/use_windows_codec_binaries: false/" ivas-processing-scripts/experiments/selection//config/.yml

		# run experiments test
		- cd ivas-processing-scripts
		- python3 -m pytest tests/test_experiments.py::test_generate_test_items -n auto \| tee log.txt
		artifacts:
		paths:
		- ivas-processing-scripts/experiments/selection//proc_output/.log
		- log.txt
		when: on_failure
		expire_in: 1 week

		# ---------------------------------------------------------------
		# Test jobs for main branch
		# ---------------------------------------------------------------

lib_com/options.h

100755 → 100644

+2 −0

Original line number	Diff line number	Diff line
		@@ -148,10 +148,12 @@
		#define FIX_558_PLC_DISCONT /* FhG: issue 558: fix discontinuities in DFT Stereo when switching from TCX concealment to ACELP */
		#define FIX_564 /* Nokia: Issue 564: Fix gains in JBM path for SBA with parametric binaural renderer */
		#define FIX_566_2DIR_MASA_384K /* Nokia: Issued 566: Bugfix in 384k MASA metadata encoding of second direction */
		#define FIX_568_ISM_BITRATE_SWITCHING /* Philips: Issue 568: Bugfix for renderer re-initialization by ISM and bitrate switching */
		#define FIX_565_SBA_BURST_IN_FEC /* VA: Issue 565: Fix noise burst during FEC, due to wrong total_brate initialization */
		#define FIX_562_ISM2_64KBPS /* VA: issue 562: fix ISM2 at 64kbps issue */
		#define FIX_559_EXTL_IGF_MISMATCH /* VA: issue 559: fix mismatch between st->extl and st->igf observed as crash in PlanarSBA bitrate switching */

		#define FIX_QMETA_SID_5k2 /* Nokia: Issue 137: enable using full 5.2k bitrate in MASA SID */

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

lib_dec/ivas_ism_dec.c

+16 −1

Original line number	Diff line number	Diff line
		@@ -168,7 +168,11 @@ static ivas_error ivas_ism_bitrate_switching(
		/* Deallocate the ParamISM struct */
		ivas_param_ism_dec_close( &( st_ivas->hDirAC ), st_ivas->hDecoderConfig->output_config );

		#ifdef FIX_568_ISM_BITRATE_SWITCHING
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL \|\| st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		#else
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL )
		#endif
		{
		/* close the parametric binaural renderer */
		ivas_dirac_dec_close_binaural_data( &st_ivas->hDiracDecBin );
		@@ -198,7 +202,11 @@ static ivas_error ivas_ism_bitrate_switching(
		}
		}

		#ifdef FIX_568_ISM_BITRATE_SWITCHING
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_IR )
		#else
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_IR \|\| st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		#endif
		{
		/* close the parametric binaural renderer */
		ivas_dirac_dec_close_binaural_data( &st_ivas->hDiracDecBin );
		@@ -221,8 +229,11 @@ static ivas_error ivas_ism_bitrate_switching(
		{
		return error;
		}

		#ifdef FIX_568_ISM_BITRATE_SWITCHING
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL \|\| st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		#else
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL )
		#endif
		{
		/* open the parametric binaural renderer */
		if ( ( error = ivas_dirac_dec_binaural_copy_hrtfs( &st_ivas->hHrtfParambin ) ) != IVAS_ERR_OK )
		@@ -260,7 +271,11 @@ static ivas_error ivas_ism_bitrate_switching(
		}
		}

		#ifdef FIX_568_ISM_BITRATE_SWITCHING
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_IR )
		#else
		if ( st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_IR \|\| st_ivas->hOutSetup.output_config == AUDIO_CONFIG_BINAURAL_ROOM_REVERB )
		#endif
		{
		/* open the parametric binaural renderer */
		if ( ( error = ivas_dirac_dec_binaural_copy_hrtfs( &st_ivas->hHrtfParambin ) ) != IVAS_ERR_OK )

lib_dec/ivas_qmetadata_dec.c

+86 −8

Original line number	Diff line number	Diff line
		@@ -1181,6 +1181,9 @@ int16_t ivas_qmetadata_dec_sid_decode(
		float avg_direction_vector[3];
		float direction_vector[3];
		int16_t metadata_sid_bits; /* bits allocated to SID for metadata */
		#ifdef FIX_QMETA_SID_5k2
		int16_t bits_delta, bits_dir;
		#endif
		#ifdef DEBUG_MODE_QMETADATA
		static FILE *pF = NULL;
		static FILE *pF_azi = NULL;
		@@ -1204,8 +1207,12 @@ int16_t ivas_qmetadata_dec_sid_decode(
		}
		else
		{
		#ifdef FIX_QMETA_SID_5k2
		metadata_sid_bits = ( IVAS_SID_5k2 - SID_2k40 ) / FRAMES_PER_SEC - SID_FORMAT_NBITS;
		#else
		/* TODO: still use old sid frame size to keep bitexactness */
		metadata_sid_bits = ( 4400 /IVAS_SID_5k2/ - SID_2k40 ) / FRAMES_PER_SEC - SID_FORMAT_NBITS;
		#endif
		}

		start_index = *index;
		@@ -1247,16 +1254,80 @@ int16_t ivas_qmetadata_dec_sid_decode(
		{
		q_direction->not_in_2D = 1;
		}

		#ifdef FIX_QMETA_SID_5k2
		bits_dir = 0;
		if ( ivas_format != SBA_FORMAT )
		{
		/* Decode diffuseness*/
		for ( b = start_band; b < nbands; b++ )
		{
		diffuseness_index[b] = ivas_qmetadata_DecodeQuasiUniform( bitstream, index, DIRAC_DIFFUSE_LEVELS - 4 ) + 4;
		q_direction->band_data[b].energy_ratio_index[0] = diffuseness_index[b];
		q_direction->band_data[b].bits_sph_idx[0] = bits_direction_masa[diffuseness_index[b]];
		bits_dir += q_direction->band_data[b].bits_sph_idx[0];
		q_direction->band_data[b].azimuth_m_alphabet[0] = no_phi_masa[q_direction->band_data[b].bits_sph_idx[0] - 1][0];
		}

		bits_delta = metadata_sid_bits - ( start_index - index ) - bits_dir; / bit_diff is already read */

		if ( bits_delta > 0 )
		{
		while ( bits_delta > 0 )
		{
		for ( b = start_band; b < nbands && ( bits_delta > 0 ); b++ )
		{
		if ( q_direction->band_data[b].bits_sph_idx[0] < 11 )
		{
		bits_delta -= 1;
		q_direction->band_data[b].bits_sph_idx[0]++;
		}
		}
		}
		if ( q_direction->not_in_2D == 0 )
		{
		for ( b = start_band; b < nbands; b++ )
		{
		q_direction->band_data[b].azimuth_m_alphabet[0] = 1 << ( min( 8, q_direction->band_data[b].bits_sph_idx[0] ) );
		}
		}
		}
		else
		{
		while ( bits_delta < 0 )
		{
		for ( b = nbands - 1; b >= start_band && ( bits_delta < 0 ); b-- )
		{
		if ( q_direction->band_data[b].bits_sph_idx[0] >= 4 )
		{
		bits_delta += 1;
		q_direction->band_data[b].bits_sph_idx[0]--;
		}
		}
		if ( q_direction->not_in_2D == 0 )
		{
		for ( b = start_band; b < nbands; b++ )
		{
		q_direction->band_data[b].azimuth_m_alphabet[0] = 1 << ( min( 8, q_direction->band_data[b].bits_sph_idx[0] ) );
		}
		}
		}
		}
		}
		else
		{
		#endif

		/* Decode diffuseness*/
		for ( b = start_band; b < nbands; b++ )
		{
		diffuseness_index[b] = ivas_qmetadata_DecodeQuasiUniform( bitstream, index, DIRAC_DIFFUSE_LEVELS - 4 ) + 4;
		q_direction->band_data[b].energy_ratio_index[0] = diffuseness_index[b];
		q_direction->band_data[b].bits_sph_idx[0] = bits_direction_masa[diffuseness_index[b]];
		q_direction->band_data[b].azimuth_m_alphabet[0] = no_phi_masa[q_direction->band_data[b].bits_sph_idx[0] - 1][0];
		}
		#ifdef FIX_QMETA_SID_5k2
		}
		#endif
		for ( b = start_band; b < nbands; b++ )
		{
		q_direction->band_data[b].energy_ratio[0] = 1.0f - diffuseness_reconstructions[diffuseness_index[b]];
		@@ -1299,7 +1370,14 @@ int16_t ivas_qmetadata_dec_sid_decode(
		{
		for ( b = start_band; b < nbands; b++ )
		{
		#ifdef FIX_QMETA_SID_5k2
		if ( ivas_format == SBA_FORMAT )
		{
		#endif
		q_direction->band_data[b].azimuth_m_alphabet[0] = 1 << ( min( 5, q_direction->band_data[b].bits_sph_idx[0] ) );
		#ifdef FIX_QMETA_SID_5k2
		}
		#endif
		q_direction->band_data[b].azimuth_index[0] = ivas_qmetadata_DecodeQuasiUniform( bitstream, index, q_direction->band_data[b].azimuth_m_alphabet[0] );
		q_direction->band_data[b].azimuth_index[0] = ivas_qmetadata_dereorder_generic( q_direction->band_data[b].azimuth_index[0] ) + ( q_direction->band_data[b].azimuth_m_alphabet[0] >> 1 );
		avg_azimuth = 360.0f / (float) ( q_direction->band_data[b].azimuth_m_alphabet[0] ) * q_direction->band_data[b].azimuth_index[0] - 180;

lib_enc/ivas_qmetadata_enc.c

+94 −18

Original line number	Diff line number	Diff line
		@@ -833,6 +833,10 @@ ivas_error ivas_qmetadata_enc_encode_hr_384_512(
		}
		#endif

		#ifdef FIX_QMETA_SID_5k2
		q_direction->not_in_2D = 0;
		#endif

		/Coherence /
		if ( all_coherence_zero == 0 )
		{
		@@ -980,7 +984,11 @@ void ivas_qmetadata_enc_sid_encode(
		else
		{
		/* TODO: still use old sid frame size to keep bitexactness */
		#ifdef FIX_QMETA_SID_5k2
		metadata_sid_bits = ( IVAS_SID_5k2 - SID_2k40 ) / FRAMES_PER_SEC - SID_FORMAT_NBITS;
		#else
		metadata_sid_bits = ( 4400 /IVAS_SID_5k2/ - SID_2k40 ) / FRAMES_PER_SEC - SID_FORMAT_NBITS;
		#endif
		}

		#ifdef DEBUG_MODE_QMETADATA
		@@ -1036,12 +1044,77 @@ void ivas_qmetadata_enc_sid_encode(
		/Encode the quantized diffuseness in raw coding/
		bits_dir = 0;
		bits_diff = 0;
		#ifdef FIX_QMETA_SID_5k2
		if ( ivas_format != SBA_FORMAT )
		{
		for ( b = start_band; b < nbands; b++ )
		{
		q_direction->band_data[b].energy_ratio_index[0] = max( q_direction->band_data[b].energy_ratio_index[0], 4 );
		bits_diff += ivas_qmetadata_encode_quasi_uniform_length( q_direction->band_data[b].energy_ratio_index[0] - 4, DIRAC_DIFFUSE_LEVELS - 4 );
		q_direction->band_data[b].bits_sph_idx[0] = bits_direction_masa[q_direction->band_data[b].energy_ratio_index[0]];

		if ( q_direction->not_in_2D == 0 )
		{
		q_direction->band_data[b].azimuth_m_alphabet[0] = 1 << ( min( 8, q_direction->band_data[b].bits_sph_idx[0] ) );
		bits_dir += ivas_qmetadata_encode_quasi_uniform_length( q_direction->band_data[b].azimuth_m_alphabet[0] - 1, q_direction->band_data[b].azimuth_m_alphabet[0] );
		}
		else
		{
		bits_dir += q_direction->band_data[b].bits_sph_idx[0];
		}
		}

		/* Reduce bit demand by increasing diffuseness*/
		bits_delta = metadata_sid_bits - ( hMetaData->nb_bits_tot - bit_pos_start ) - bits_diff - bits_dir;
		if ( bits_delta > 0 )
		{
		while ( bits_delta > 0 )
		{
		for ( b = start_band; b < nbands && ( bits_delta > 0 ); b++ )
		{
		if ( q_direction->band_data[b].bits_sph_idx[0] < 11 )
		{
		bits_delta -= 1;
		q_direction->band_data[b].bits_sph_idx[0]++;
		}
		}
		}
		if ( q_direction->not_in_2D == 0 )
		{
		for ( b = start_band; b < nbands; b++ )
		{
		q_direction->band_data[b].azimuth_m_alphabet[0] = 1 << ( min( 8, q_direction->band_data[b].bits_sph_idx[0] ) );
		}
		}
		}
		else
		{
		while ( bits_delta < 0 )
		{
		for ( b = nbands - 1; b >= start_band && ( bits_delta < 0 ); b-- )
		{
		if ( q_direction->band_data[b].bits_sph_idx[0] >= 4 )
		{
		bits_delta += 1;
		q_direction->band_data[b].bits_sph_idx[0]--;
		if ( q_direction->not_in_2D == 0 )
		{
		q_direction->band_data[b].azimuth_m_alphabet[0] = 1 << ( min( 8, q_direction->band_data[b].bits_sph_idx[0] ) );
		}
		}
		}
		}
		}
		}
		else
		{
		#endif
		for ( b = start_band; b < nbands; b++ )
		{
		q_direction->band_data[b].energy_ratio_index[0] = max( q_direction->band_data[b].energy_ratio_index[0], 4 );
		bits_diff += ivas_qmetadata_encode_quasi_uniform_length( q_direction->band_data[b].energy_ratio_index[0] - 4, DIRAC_DIFFUSE_LEVELS - 4 );
		q_direction->band_data[b].bits_sph_idx[0] = bits_direction_masa[q_direction->band_data[b].energy_ratio_index[0]];

		if ( q_direction->not_in_2D == 0 )
		{
		q_direction->band_data[b].azimuth_m_alphabet[0] = 1 << ( min( 5, q_direction->band_data[b].bits_sph_idx[0] ) );
		@@ -1069,6 +1142,9 @@ void ivas_qmetadata_enc_sid_encode(
		}
		}
		}
		#ifdef FIX_QMETA_SID_5k2
		}
		#endif
		assert( ( bits_delta >= 0 ) && "Bit budget in Qmetadata SID is violated!!!" );

		/Code diffuseness/