Merge branch 'Ericsson/TD-renderer-ITD-improvement-and-cleanup' into 'main' (4353d954) · Commits · IVAS Codec Public Collaboration / IVAS Codec

.gitlab-ci.yml

+6 −4

Original line number	Diff line number	Diff line
		@@ -331,7 +331,7 @@ external-renderer-make-pytest:
		- make -j IVAS_rend
		- make -j unittests
		- make -j --directory scripts/td_object_renderer/object_renderer_standalone
		- python3 -m pytest -q --log-level ERROR -n auto -rA --junit-xml=report-junit.xml tests/renderer/test_renderer.py
		- python3 -m pytest -q -n auto -rA --junit-xml=report-junit.xml tests/renderer/test_renderer.py
		artifacts:
		name: "mr-$CI_MERGE_REQUEST_IID--sha-$CI_COMMIT_SHORT_SHA--job-$CI_JOB_NAME--results"
		when: always
		@@ -353,7 +353,8 @@ external-renderer-cmake-asan-pytest:
		- python3 ci/disable_ram_counting.py
		- cmake -B cmake-build -G "Unix Makefiles" -DCLANG=asan -DCOPY_EXECUTABLES_FROM_BUILD_DIR=true
		- cmake --build cmake-build -- -j
		- python3 -m pytest -q --log-level ERROR -n auto -rA --junit-xml=report-junit.xml tests/renderer/test_renderer.py
		- python3 -m pytest -q -n auto -rA --junit-xml=report-junit.xml tests/renderer/test_renderer.py

		artifacts:
		name: "mr-$CI_MERGE_REQUEST_IID--sha-$CI_COMMIT_SHORT_SHA--job-$CI_JOB_NAME--results"
		when: always
		@@ -375,7 +376,8 @@ external-renderer-cmake-msan-pytest:
		- python3 ci/disable_ram_counting.py
		- cmake -B cmake-build -G "Unix Makefiles" -DCLANG=msan -DCOPY_EXECUTABLES_FROM_BUILD_DIR=true
		- cmake --build cmake-build -- -j
		- python3 -m pytest -q --log-level ERROR -n auto -rA --junit-xml=report-junit.xml tests/renderer/test_renderer.py
		- python3 -m pytest -q -n auto -rA --junit-xml=report-junit.xml tests/renderer/test_renderer.py

		artifacts:
		name: "mr-$CI_MERGE_REQUEST_IID--sha-$CI_COMMIT_SHORT_SHA--job-$CI_JOB_NAME--results"
		when: always
		@@ -397,7 +399,7 @@ external-renderer-cmake-msan-pytest:
		script:
		- cmake -B cmake-build -G "Unix Makefiles" -DCOPY_EXECUTABLES_FROM_BUILD_DIR=true -DDEC_TO_REND_FLOAT_DUMP=true
		- cmake --build cmake-build -- -j
		- python3 -m pytest -q --log-level ERROR -n 1 -rA --junit-xml=report-junit.xml tests/renderer/test_renderer_vs_decoder.py
		- python3 -m pytest -q -n 1 -rA --junit-xml=report-junit.xml tests/renderer/test_renderer_vs_decoder.py
		artifacts:
		name: "mr-$CI_MERGE_REQUEST_IID--sha-$CI_COMMIT_SHORT_SHA--job-$CI_JOB_NAME--results"
		when: always

lib_com/ivas_cnst.h

+12 −1

Original line number	Diff line number	Diff line
		@@ -1388,8 +1388,19 @@ typedef enum
		#define SFX_SPAT_BIN_MAX_FILTER_LENGTH 256

		#define SPAT_BIN_MAX_INPUT_CHANNELS 1 /* Max number of input channels per source/object. Mono for now, but stereo objects may be considered. */

		#ifdef FIX_ITD
		#define MAX_ITD 50
		#define SFX_SPAT_BIN_SINC_M 5
		#define SFX_SPAT_BIN_NUM_SUBSAMPLES 64
		#define ITD_MEM_LEN (MAX_ITD + SFX_SPAT_BIN_SINC_M)
		#define L_SUBFRAME5MS_48k (L_FRAME48k/4)
		#define MAX_ANGULAR_STEP (15.0f)
		#define MAX_ANGULAR_STEP_INV ( 1.0f / MAX_ANGULAR_STEP )
		#define MAX_INTERPOLATION_STEPS 12
		#define BINAURAL_TD_LATENCY_S 0.0f /* ITD fix removes TD renderer delay -- should be cleaned out */
		#else
		#define BINAURAL_TD_LATENCY_S 0.00675f /* Binaural TD renderer latency in second == 324 samples in between 333 and 315 */
		#endif

		/* ----- Enums - TD Renderer ----- */

lib_com/ivas_prot.h

+60 −2

Original line number	Diff line number	Diff line
		@@ -5004,7 +5004,13 @@ void GetFilterFromAngle(
		TDREND_HRFILT_FiltSet_t HrFiltSet_p, / i/o: HR filter set structure */
		const float Elev, /* i : Elevation, degrees */
		float Azim, /* i : Azimuth, degrees */
		#ifdef FIX_ITD
		float LeftFilter, / o : Left HR filter */
		float RightFilter, / o : Right HR filter */
		int16_t itd / o : ITD value */
		#else
		SFX_SpatBin_Params_t SfxSpatBinParams_p / i/o: Currently used HR filter */
		#endif
		);

		void HRTF_model_precalc(
		@@ -5029,12 +5035,22 @@ void TDREND_HRFILT_SetFiltSet(

		ivas_error TDREND_REND_RenderSourceHRFilt(
		TDREND_SRC_t Src_p, / i/o: The source to be rendered */
		#ifdef FIX_ITD
		const float hrf_left_delta, / i: Left filter interpolation delta */
		const float hrf_right_delta, / i: Right filter interpolation delta */
		const int16_t intp_count, /* i: Interpolation count */
		#else
		#ifdef TDREND_HRTF_TABLE_METHODS
		BINAURAL_TD_OBJECT_RENDERER_HANDLE hBinRendererTd, /* i/o: TD renderer handle */
		#endif
		#endif
		float output_buf[][L_SPATIAL_SUBFR_48k], /* o : Output buffer */
		#ifdef FIX_ITD
		const int16_t subframe_length /* i : Subframe length in use */
		#else
		const int16_t subframe_length, /* i : Subframe length in use */
		const int32_t output_Fs /* i : Output sample rate */
		#endif
		);

		/* ----- Object renderer - sources ----- */
		@@ -5067,7 +5083,20 @@ void TDREND_SRC_REND_UpdateFiltersFromSpatialParams(
		BINAURAL_TD_OBJECT_RENDERER_HANDLE hBinRendererTd, /* i/o: TD renderer handle */
		TDREND_SRC_REND_t SrcRend_p, / i/o: Source object */
		TDREND_SRC_SPATIAL_t SrcSpatial_p, / i : Spatial aspects of source */
		#ifdef FIX_ITD
		float hrf_left_prev, / o: Left filter */
		float hrf_right_prev, / o: Right filter */
		float hrf_left_delta, / o: Left filter interpolation delta */
		float hrf_right_delta, / o: Right filter interpolation delta */
		int16_t intp_count, / o: Interpolation count */
		int16_t filterlength, / o: Length of filters */
		int16_t itd, / o: ITD value */
		float Gain, / o: Gain value */
		TDREND_SRC_t *Src_p,
		const int16_t subframe_idx /* i : Subframe index to 5 ms subframe */
		#else
		const int32_t output_Fs /* i : Output sample rate */
		#endif
		);

		ivas_error TDREND_SRC_Alloc(
		@@ -5080,8 +5109,12 @@ void TDREND_SRC_Dealloc(

		void TDREND_SRC_Init(
		TDREND_SRC_t Src_p, / i/o: Source to initialize */
		#ifdef FIX_ITD
		const TDREND_PosType_t PosType /* i : Position type specifier */
		#else
		const TDREND_PosType_t PosType, /* i : Position type specifier */
		const int32_t output_Fs /* i : Output sampling rate */
		#endif
		);

		/* ----- Object renderer - vec ----- */
		@@ -5126,8 +5159,12 @@ int16_t TDREND_SPATIAL_EvalOrthonormOrient(
		ivas_error TDREND_MIX_AddSrc(
		BINAURAL_TD_OBJECT_RENDERER_HANDLE hBinRendererTd, /* i/o: TD renderer handle */
		int16_t SrcInd, / o : Source index */
		#ifdef FIX_ITD
		const TDREND_PosType_t PosType /* i : Position type (absolute/relative) */
		#else
		const TDREND_PosType_t PosType, /* i : Position type (absolute/relative) */
		const int32_t output_Fs /* i : Output sampling rate */
		#endif
		);

		ivas_error TDREND_MIX_SetDistAttenModel(
		@@ -5158,7 +5195,7 @@ ivas_error TDREND_MIX_Init(
		);

		/* ----- Object renderer - sfx ----- */

		#ifndef FIX_ITD
		ivas_error TDREND_SFX_SpatBin_Initialize(
		SFX_SpatBin_t SfxSpatBin_p, / i/o: Spatial parameters handle */
		const int32_t output_Fs /* i : Output sampling rate */
		@@ -5180,8 +5217,29 @@ void TDREND_SFX_SpatBin_Execute_Main(
		int16_t NoOfDeliveredOutputSamples_p, / o : Number of output samples actually rendered */
		const int32_t output_Fs /* i : Output sample rate */
		);
		#endif


		#ifdef FIX_ITD
		void TDREND_Apply_ITD(
		float input, / i: Input SCE subframe to be time adjusted */
		float out_left, / o: Output left channels with ITD applied */
		float out_right, / o: Output right channels with ITD applied */
		int16_t previtd, /i/o: Previous ITD value */
		const int16_t itd, /* i: Current subframe ITD value */
		float mem_itd, /i/o: ITD buffer memory */
		const int16_t length /* i: Subframe length */
		);

		void TDREND_firfilt(
		float signal, / i/o: Input signal / Filtered signal */
		float filter, / i/o: FIR filter */
		const float filter_delta, / i : FIR filter delta */
		const int16_t intp_count, /* i : interpolation count */
		float mem, / i/o: filter memory */
		const int16_t subframe_length, /* i : Length of signal */
		const int16_t filterlength /* i : Filter length */
		);
		#endif
		/----------------------------------------------------------------------------------
		* Filter-bank (FB) Mixer
		----------------------------------------------------------------------------------/

lib_com/options.h

100755 → 100644

+1 −0

Original line number	Diff line number	Diff line
		@@ -153,6 +153,7 @@
		#define FIX_MCT_PLC_RECOVERY /* Issue 184: scale the old synthesis part correctly in the first good frame after lost frames in MCT modes - to be activated after previous switch is merged */
		#define SBA_BR_SWITCHING /* Issue 114: Changes for sba bit rate switching*/
		#define FIX_AGC_WINFUNC_MEMORY /* Issue 62: lower agc_com.winFunc memory consumption */
		#define FIX_ITD /* Contribution 16: TD renderer ITD improvement and code cleanup */
		#define REMOVE_SID_HARM_LEFTOVERS /* Issue 192: remove leftovers from the SID bitrate harmonization */
		#define FIX_MCT_UNINIT_MEM /* Issue 166: Reading of uninitialized memory in TCX range coder */
		#define FIX_IGF_NOISE_REPETITION /* Issue 182: fix repetition of same noise in IGF */

lib_dec/ivas_stat_dec.h

+18 −3

Original line number	Diff line number	Diff line
		@@ -1328,6 +1328,8 @@ typedef struct ivas_binaural_head_track_struct
		/----------------------------------------------------------------------------------
		* TD ISm Object Renderer structure
		----------------------------------------------------------------------------------/

		#ifndef FIX_ITD
		// VE2AT: move to ivas_rom_rend.h ?
		typedef struct
		{
		@@ -1391,6 +1393,7 @@ typedef struct TDREND_LIST_Item_s
		struct TDREND_LIST_Item_s Next_p; / Pointer to the next item */

		} TDREND_LIST_Item_t;
		#endif

		typedef struct
		{
		@@ -1583,10 +1586,10 @@ typedef struct TDREND_SRC_REND_s
		float SrcGainMax_p[SPAT_BIN_MAX_INPUT_CHANNELS];
		float DirGain_p[SPAT_BIN_MAX_INPUT_CHANNELS];
		float DistGain_p[SPAT_BIN_MAX_INPUT_CHANNELS];

		#ifndef FIX_ITD
		/* HR filtering parameters */
		SFX_SpatBin_t *SfxSpatBin_p;

		#endif
		} TDREND_SRC_REND_t;


		@@ -1609,7 +1612,19 @@ typedef struct
		float InputFrame_p; / Input frame pointer */
		TDREND_SRC_SPATIAL_t *SrcSpatial_p;
		TDREND_SRC_REND_t *SrcRend_p;

		#ifdef FIX_ITD
		int16_t itd;
		int16_t previtd;
		int16_t filterlength;
		float mem_itd[ITD_MEM_LEN];
		float hrf_left_prev[SFX_SPAT_BIN_MAX_FILTER_LENGTH]; /* Todo: Should we allocate these buffers with count_malloc instead of the maximum length? */
		float hrf_right_prev[SFX_SPAT_BIN_MAX_FILTER_LENGTH];
		float azim_prev;
		float elev_prev;
		float mem_hrf_left[SFX_SPAT_BIN_MAX_FILTER_LENGTH - 1];
		float mem_hrf_right[SFX_SPAT_BIN_MAX_FILTER_LENGTH - 1];
		float Gain;
		#endif
		} TDREND_SRC_t;

		/* Top level TD binaural renderer handle */