Merge branch 'main' into ci/harmonize-basop-ci-branch-w-main (77413102) · Commits · IVAS Codec Public Collaboration / IVAS Codec

lib_com/options.h

100755 → 100644

+1 −0

Original line number	Diff line number	Diff line
		@@ -163,6 +163,7 @@
		#define IVAS_RTPDUMP /* RTPDUMP writing and reading for IVAS payloads */
		#define IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT /* RTPDUMP acoustic environment */
		#define FIXED_RTP_SEQUENCE_NUM /* Remove random sequence number initialization */
		#define ISM_PI_DATA /* Add reading and packing/unpacking of ISM PI data */

		/* ################### Start BE switches ################################# */
		/* only BE switches wrt selection floating point code */

lib_util/ivas_rtp_file.c

+103 −1

Original line number	Diff line number	Diff line
		@@ -152,13 +152,16 @@ void IVAS_RTP_LogPiData(
		uint32_t nPiDataPresent /* i : Number of valid elements in the piData array */
		)
		{
		#ifdef ISM_PI_DATA
		uint16_t n;
		#endif
		uint32_t timestamp = ~0u;
		if ( f_piDataOut == NULL \|\| piData == NULL \|\| nPiDataPresent == 0 )
		{
		return;
		}

		#ifdef IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT
		#if defined IVAS_RTPDUMP_ACOUSTIC_ENVIRONMENT \|\| defined ISM_PI_DATA
		#ifdef _WIN32
		if ( ftell( f_piDataOut ) > 3 )
		#else
		@@ -332,6 +335,104 @@ void IVAS_RTP_LogPiData(
		fprintf( f_piDataOut, "{}" );
		}
		break;
		#ifdef ISM_PI_DATA
		case IVAS_PI_ISM_ORIENTATION:
		{
		fprintf( f_piDataOut, "[\n" );
		for ( n = 0; n < cur->data.ismOrientation.numObjects; n++ )
		{
		if ( n != 0 )
		{
		fprintf( f_piDataOut, ",\n" );
		}
		fprintf( f_piDataOut, "\t\t\t{\n\t\t\t\t\"w\": %f,\n\t\t\t\t\"x\": %f,\n\t\t\t\t\"y\": %f,\n\t\t\t\t\"z\": %f \n\t\t\t}",
		cur->data.ismOrientation.orientation[n].w, cur->data.ismOrientation.orientation[n].x, cur->data.ismOrientation.orientation[n].y, cur->data.ismOrientation.orientation[n].z );
		}
		fprintf( f_piDataOut, "\n\t\t]" );
		}
		break;
		case IVAS_PI_ISM_NUM:
		{
		fprintf( f_piDataOut, "{\n" );
		fprintf( f_piDataOut, "\t\t\t\"num\": %d", cur->data.ismNum.numObjects );
		fprintf( f_piDataOut, "\n\t\t}" );
		}
		break;
		case IVAS_PI_ISM_ID:
		{
		fprintf( f_piDataOut, "{\n" );
		fprintf( f_piDataOut, "\t\t\t\"ids\": [\n" );
		for ( n = 0; n < cur->data.ismId.numObjects; n++ )
		{
		if ( n != 0 )
		{
		fprintf( f_piDataOut, ",\n" );
		}
		fprintf( f_piDataOut, "\t\t\t\t%d", cur->data.ismId.id[n] );
		}
		fprintf( f_piDataOut, "\n\t\t\t]" );
		fprintf( f_piDataOut, "\n\t\t}" );
		}
		break;
		case IVAS_PI_ISM_GAIN:
		{
		fprintf( f_piDataOut, "{\n" );
		fprintf( f_piDataOut, "\t\t\t\"gains\": [\n" );
		for ( n = 0; n < cur->data.ismGain.numObjects; n++ )
		{
		if ( n != 0 )
		{
		fprintf( f_piDataOut, ",\n" );
		}
		fprintf( f_piDataOut, "\t\t\t\t%d", cur->data.ismGain.dB[n] );
		}
		fprintf( f_piDataOut, "\n\t\t\t]" );
		fprintf( f_piDataOut, "\n\t\t}" );
		}
		break;
		case IVAS_PI_ISM_POSITION:
		{
		fprintf( f_piDataOut, "[\n" );
		for ( n = 0; n < cur->data.ismPosition.numObjects; n++ )
		{
		if ( n != 0 )
		{
		fprintf( f_piDataOut, ",\n" );
		}
		fprintf( f_piDataOut, "\t\t\t{\n\t\t\t\t\"x\": %f,\n\t\t\t\t\"y\": %f,\n\t\t\t\t\"z\": %f \n\t\t\t}", cur->data.ismPosition.position[n].x, cur->data.ismPosition.position[n].y, cur->data.ismPosition.position[n].z );
		}
		fprintf( f_piDataOut, "\n\t\t]" );
		}
		break;
		case IVAS_PI_ISM_DISTANCE_ATTENUATION:
		{
		fprintf( f_piDataOut, "[\n" );
		for ( n = 0; n < cur->data.ismAttenuation.numObjects; n++ )
		{
		if ( n != 0 )
		{
		fprintf( f_piDataOut, ",\n" );
		}
		fprintf( f_piDataOut, "\t\t\t{\n\t\t\t\t\"ref_dist\": %f,\n\t\t\t\t\"max_dist\": %f,\n\t\t\t\t\"roll_off\": %f \n\t\t\t}", cur->data.ismAttenuation.distAtten[n].ref_dist, cur->data.ismAttenuation.distAtten[n].max_dist, cur->data.ismAttenuation.distAtten[n].roll );
		}
		fprintf( f_piDataOut, "\n\t\t]" );
		}
		break;
		case IVAS_PI_ISM_DIRECTIVITY:
		{
		fprintf( f_piDataOut, "[\n" );
		for ( n = 0; n < cur->data.ismDirectivity.numObjects; n++ )
		{
		if ( n != 0 )
		{
		fprintf( f_piDataOut, ",\n" );
		}
		fprintf( f_piDataOut, "\t\t\t{\n\t\t\t\t\"inner_ang\": %d,\n\t\t\t\t\"outer_ang\": %d,\n\t\t\t\t\"outer_att\": %f \n\t\t\t}", cur->data.ismDirectivity.directivity[n].innerConeAngle, cur->data.ismDirectivity.directivity[n].outerConeAngle, cur->data.ismDirectivity.directivity[n].outerAttenuationdB );
		}
		fprintf( f_piDataOut, "\n\t\t]" );
		}
		break;
		#else
		case IVAS_PI_ISM_NUM:
		case IVAS_PI_ISM_ID:
		case IVAS_PI_ISM_GAIN:
		@@ -339,6 +440,7 @@ void IVAS_RTP_LogPiData(
		case IVAS_PI_ISM_POSITION:
		case IVAS_PI_ISM_DISTANCE_ATTENUATION:
		case IVAS_PI_ISM_DIRECTIVITY:
		#endif
		case IVAS_PI_PI_LATENCY:
		case IVAS_PI_R_ISM_ID:
		case IVAS_PI_R_ISM_GAIN:

lib_util/ivas_rtp_pi_data.c

+542 −9

File changed.

Preview size limit exceeded, changes collapsed.

lib_util/ivas_rtp_pi_data.h

+30 −13

Original line number	Diff line number	Diff line
		@@ -216,6 +216,9 @@ typedef struct
		{
		size_t size; /* sizeof(IVAS_PIDATA_ISM_ID) */
		uint32_t piDataType; /* IVAS_PI_ISM_ID */
		#ifdef ISM_PI_DATA
		uint16_t numObjects; /* number of objects */
		#endif
		uint8_t id[IVAS_MAX_NUM_OBJECTS]; /* 8-bit ISM id of object */
		} IVAS_PIDATA_ISM_ID;

		@@ -224,6 +227,9 @@ typedef struct
		{
		size_t size; /* sizeof(IVAS_PIDATA_ISM_GAIN) */
		uint32_t piDataType; /* IVAS_PI_ISM_GAIN */
		#ifdef ISM_PI_DATA
		uint16_t numObjects; /* number of objects */
		#endif
		int8_t dB[IVAS_MAX_NUM_OBJECTS]; /* ISM gain in dB per object [-96, +3] */
		} IVAS_PIDATA_ISM_GAIN;

		@@ -232,6 +238,9 @@ typedef struct
		{
		size_t size; /* sizeof(IVAS_PIDATA_ISM_ORIENTATION) */
		uint32_t piDataType; /* IVAS_PI_ISM_ORIENTATION */
		#ifdef ISM_PI_DATA
		uint16_t numObjects; /* number of objects */
		#endif
		IVAS_QUATERNION orientation[IVAS_MAX_NUM_OBJECTS]; /* Orientation of audio objects in ISM(s) */
		} IVAS_PIDATA_ISM_ORIENTATION;

		@@ -240,6 +249,9 @@ typedef struct
		{
		size_t size; /* sizeof(IVAS_PIDATA_ISM_POSITION) */
		uint32_t piDataType; /* IVAS_PI_ISM_POSITION */
		#ifdef ISM_PI_DATA
		uint16_t numObjects; /* number of objects */
		#endif
		IVAS_COORDINATE position[IVAS_MAX_NUM_OBJECTS]; /* Position of audio objects in ISM(s) */
		} IVAS_PIDATA_ISM_POSITION;

		@@ -259,6 +271,9 @@ typedef struct
		{
		size_t size; /* sizeof(IVAS_PIDATA_ISM_ATTENUATION) */
		uint32_t piDataType; /* IVAS_PI_ISM_DISTANCE_ATTENUATION */
		#ifdef ISM_PI_DATA
		uint16_t numObjects; /* number of objects */
		#endif
		IVAS_DIST_ATTEN distAtten[IVAS_MAX_NUM_OBJECTS]; /* Distance attenuation of audio objects */
		} IVAS_PIDATA_ISM_ATTENUATION;

		@@ -278,6 +293,9 @@ typedef struct
		{
		size_t size; /* sizeof(IVAS_PIDATA_ISM_DIRECTIVITY) */
		uint32_t piDataType; /* IVAS_PI_ISM_DIRECTIVITY */
		#ifdef ISM_PI_DATA
		uint16_t numObjects; /* number of objects */
		#endif
		IVAS_ISM_DIRECTIVITY directivity[IVAS_MAX_NUM_OBJECTS]; /* Directivity of audio objects */
		} IVAS_PIDATA_ISM_DIRECTIVITY;

		@@ -337,7 +355,6 @@ typedef struct
		IVAS_COORDINATE position; /* Position of audio objects in ISM(s) */
		} IVAS_PIDATA_LISTENER_POSITION;


		/* Dynamic Audio Suppression describes receiver’s preference with respect to the
		* type of audio content that should be enhanced and the amount of suppression to
		* be applied to the background noise

tests/rtp/ivasrtp.py

+154 −11

Original line number	Diff line number	Diff line
		@@ -389,6 +389,46 @@ class ACOUSTIC_ENVIRONMENT:
		abscoeff: tuple[float, float, float, float, float, float] = ()


		@dataclass
		class ISM_NUM:
		num: int = 1

		@dataclass
		class ISM_ID:
		ids: list[int]

		@dataclass
		class ISM_GAIN:
		gains: list[int]

		@dataclass
		class ISM_ORIENTATION:
		orientations: list[ORIENTATION]

		@dataclass
		class ISM_POSITION:
		positions: list[POSITION]

		@dataclass
		class DISTANCE_ATTENUATION:
		ref_dist: float = 1.0
		max_dist: int = 10
		roll_off: float = 1.0

		@dataclass
		class ISM_DISTANCE_ATTENUATION:
		distance_attenuations: list[DISTANCE_ATTENUATION]

		@dataclass
		class DIRECTIVITY:
		inner_ang: int = 360
		outer_ang: int = 0
		outer_att: int = 0

		@dataclass
		class ISM_DIRECTIVITY:
		directivities: list[DIRECTIVITY]

		@dataclass
		class AUDIO_FOCUS:
		direction: Optional[ORIENTATION] = None
		@@ -402,7 +442,7 @@ class PIDATA:
		data: any = None


		MAX_PACKED_PI_SIZE = 32
		MAX_PACKED_PI_SIZE = 48
		ivasBitrates = [
		13200,
		16400,
		@@ -581,6 +621,12 @@ roomDimensionValue = [
		90.51,
		]
		absorptionCoeffValues = [0.0800, 0.1656, 0.3430, 0.7101]
		ismGains = list(range(0, -97, -1)) + [-float("inf")] + list(range(1, 4))
		refDistances = [round(x * 0.1, 1) for x in range(1, 65)]
		maxDistances = list(range(1,65))
		rolloffFactors = [round(x * 0.1, 1) for x in range(0, 41)]
		innerOuterAngles = list(range(0,361,15))
		outerAttenuations = [-float("inf")] + list(range(-90,1,3))
		codedFormats = list(FORMATS)
		codedSubFormats = list(SUBFORMATS)
		PiTypeNames = list(PIDATAS)
		@@ -923,6 +969,103 @@ def packAudioFocus(bitstrm: BitStream, data: any):
		bitstrm.append(f"uint:4={level}")
		bitstrm.append(f"uint:4=0")

		def unpackISMNum(bitstrm: ConstBitStream, piSize: int) -> ISM_NUM:
		assert piSize == 1, "Incorrect PI Data Size for ISM_NUM"
		numISM = bitstrm.read(2).uint + 1
		bitstrm.bytealign()
		return ISM_NUM(num=numISM)

		def packISMNum(bitstrm: BitStream, data: any):
		assert type(data) == ISM_NUM, "Data of type ISM_NUM is expected"
		ism_num = cast(ISM_NUM, data)
		assert ism_num.num <= 4, "Maximum 4 objects"
		bitstrm.append(f'uint:2={ism_num.num-1}')
		bitstrm.append(f'uint:6=0')

		def unpackISMID(bitstrm: ConstBitStream, piSize: int) -> ISM_ID:
		assert piSize == 1 or piSize == 2 or piSize == 3 or piSize == 4, "Incorrect PI Data Size for ISM_ID"
		IsmID = list()
		for _ in range(piSize):
		IsmID.append(bitstrm.read(8).uint+1)
		return ISM_ID(ids=IsmID)

		def packISMID(bitstrm: BitStream, data: any):
		assert type(data) == ISM_ID, "Data of type ISM_ID is expected"
		ism_id = cast(ISM_ID, data)
		assert len(ism_id.ids) <= 4, "Maximum 4 objects"
		for id in ism_id.ids:
		bitstrm.append(f'uint:8={id}')

		def unpackISMGain(bitstrm: ConstBitStream, piSize: int) -> ISM_GAIN:
		assert piSize == 1 or piSize == 2 or piSize == 3 or piSize == 4, "Incorrect PI Data Size for ISM_GAIN"
		IsmGain = list()
		for _ in range(piSize):
		IsmGain.append(bitstrm.read(7).uint)
		bitstrm.bytealign()
		return ISM_GAIN(gains=IsmGain)

		def packISMGain(bitstrm: BitStream, data: any):
		assert type(data) == ISM_GAIN, "Data of type ISM_GAIN is expected"
		ism_gain = cast(ISM_GAIN, data)
		assert len(ism_gain.gains) <= 4, "Maximum 4 objects"
		for gain in ism_gain.gains:
		gain_idx = getListIndex(ismGains, gain)
		bitstrm.append(f'uint:7={gain_idx}')
		bitstrm.append(f'uint:1=0')

		def unpackISMDistanceAttenuation(bitstrm: ConstBitStream, piSize: int) -> ISM_DISTANCE_ATTENUATION:
		ref_dist = None
		max_dist = None
		roll_off = None
		assert piSize == 13 or piSize == 23 or piSize == 33 or piSize == 43, "Incorrect PI Data Size for ISM_DISTANCE_ATTENUATION"
		ism_distance_attenutation = list()
		for _ in range(piSize):
		ref_dist = bitstrm.read(6).uint
		max_dist = bitstrm.read(6).uint
		roll_off = bitstrm.read(6).uint
		ism_distance_attenutation.append(DISTANCE_ATTENUATION(ref_dist=ref_dist, max_dist=max_dist, roll_off=roll_off))
		bitstrm.bytealign()
		return ISM_DISTANCE_ATTENUATION(distance_attenuations=ism_distance_attenutation)

		def packISMDistanceAttenuation(bitstrm: BitStream, data: any):
		assert type(data) == list, "Data of type list is expected"

		for att in cast(list, data):
		assert type(att) == DISTANCE_ATTENUATION, "Data of type list[DISTANCE_ATTENUATION] is expected"
		ref_dist_idx = getListIndex(refDistances, att.ref_dist)
		max_dist_idx = getListIndex(maxDistances, att.max_dist)
		roll_off_idx = getListIndex(rolloffFactors, att.roll_off)
		bitstrm.append(f'uint:6={ref_dist_idx}')
		bitstrm.append(f'uint:6={max_dist_idx}')
		bitstrm.append(f'uint:6={roll_off_idx}')
		bitstrm.append(f'uint:6=0')

		def unpackISMDirectivity(bitstrm: ConstBitStream, piSize: int) -> list[DISTANCE_ATTENUATION]:
		inner_ang = None
		outer_ang = None
		outer_att = None
		assert piSize == 12 or piSize == 22 or piSize == 32 or piSize == 42, "Incorrect PI Data Size for ISM_DISTANCE_ATTENUATION"
		directivities = list()
		for _ in range(piSize):
		inner_ang = bitstrm.read(5).uint
		outer_ang = bitstrm.read(5).uint
		outer_att = bitstrm.read(5).uint
		directivities.append(DIRECTIVITY(inner_ang=inner_ang, outer_ang=outer_ang, outer_att=outer_att))
		bitstrm.bytealign()
		return directivities

		def packISMDirectivity(bitstrm: BitStream, data: any):
		assert type(data) == list, "Data of type ISM_DIRECTIVITY is expected"

		for dir in cast(list, data):
		assert type(dir) == DIRECTIVITY, "Orientation PI Data expects a data of type list[DIRECTIVITY]"
		inner_ang_idx = getListIndex(innerOuterAngles, dir.inner_ang)
		outer_ang_idx = getListIndex(innerOuterAngles, dir.outer_ang)
		outer_att_idx = getListIndex(outerAttenuations, dir.outer_att)
		bitstrm.append(f'uint:5={inner_ang_idx}')
		bitstrm.append(f'uint:5={outer_ang_idx}')
		bitstrm.append(f'uint:5={outer_att_idx}')
		bitstrm.append(f'uint:1=0')

		PIDataUnpacker = [
		unpackOrientation, # SCENE_ORIENTATION,
		@@ -930,13 +1073,13 @@ PIDataUnpacker = [
		unpackOrientation, # DEVICE_ORIENTATION_UNCOMPENSATED
		unpackAcousticEnv, # ACOUSTIC_ENVIRONMENT
		unpackAudioDescription, # AUDIO_DESCRIPTION
		unpackUnsupported, # ISM_NUM
		unpackUnsupported, # ISM_ID
		unpackUnsupported, # ISM_GAIN
		unpackISMNum, # ISM_NUM
		unpackISMID, # ISM_ID
		unpackISMGain, # ISM_GAIN
		unpackOrientations, # ISM_ORIENTATION
		unpackPositions, # ISM_POSITION
		unpackUnsupported, # ISM_DISTANCE_ATTENUATION
		unpackUnsupported, # ISM_DIRECTIVITY
		unpackISMDistanceAttenuation, # ISM_DISTANCE_ATTENUATION
		unpackISMDirectivity, # ISM_DIRECTIVITY
		unpackDiegetic, # DIEGETIC_TYPE
		unpackDAS, # DYNAMIC_AUDIO_SUPPRESSION_INDICATION
		unpackAudioFocus, # AUDIO_FOCUS_INDICATION
		@@ -965,13 +1108,13 @@ PIDataPacker = [
		packOrientation, # DEVICE_ORIENTATION_UNCOMPENSATED
		packAcousticEnv, # ACOUSTIC_ENVIRONMENT
		packAudioDescription, # AUDIO_DESCRIPTION
		packUnsupported, # ISM_NUM
		packUnsupported, # ISM_ID
		packUnsupported, # ISM_GAIN
		packISMNum, # ISM_NUM
		packISMID, # ISM_ID
		packISMGain, # ISM_GAIN
		packOrientations, # ISM_ORIENTATION
		packPositions, # ISM_POSITION
		packUnsupported, # ISM_DISTANCE_ATTENUATION
		packUnsupported, # ISM_DIRECTIVITY
		packISMDistanceAttenuation, # ISM_DISTANCE_ATTENUATION
		packISMDirectivity, # ISM_DIRECTIVITY
		packDiegetic, # DIEGETIC_TYPE
		packDAS, # DYNAMIC_AUDIO_SUPPRESSION_INDICATION
		packAudioFocus, # AUDIO_FOCUS_INDICATION