Script update to generate acoustic environments metadata including early reflections (fde10181) · Commits · IVAS Codec Public Collaboration / IVAS Codec

scripts/reverb/generate_reverb_metadata.py→scripts/reverb/generate_acoustic_environments_metadata.py

+218 −7

Original line number	Diff line number	Diff line
		@@ -196,10 +196,8 @@ def get_frequency_code(f):
		def get_frequency_hop_code(index):
		assert 0 <= index < 9
		return [
		'1100', # 2^(1/8)
		'1101', # 2^(1/7)
		'0010', # 2^(1/6)
		'0011', # 2^(1/5)
		'0010', # 2^(1/12)
		'0011', # 2^(1/6)
		'0000', # 2^(1/4)
		'01', # 2^(1/3)
		'0001', # 2^(1/2)
		@@ -241,6 +239,93 @@ class fgdMethod(Enum):
		Default_Banding = '10'


		def get_distance_code(dist, isSmallScene = True):
		# 0, 1, ... 99
		metersCode = [
		'111101', '110010', '110011', '110000', '110001', '110110', '110111', '110100', '110101', '001010',
		'001011', '001000', '001001', '001110', '001111', '001100', '001101', '000010', '000011', '000000',
		'000001', '000110', '000111', '000100', '000101', '011010', '011011', '011000', '011001', '011110',
		'011111', '011100', '011101', '010010', '010011', '010000', '010001', '010110', '010111', '010100',
		'010101', '101010', '101011', '101000', '101001', '101110', '101111', '101100', '101101', '10000',
		'1000100', '1000101', '10001110', '10001111', '10001100', '10001101', '10011010', '10011011', '10011000', '10011001',
		'10011110', '10011111', '10011100', '10011101', '10010010', '10010011', '10010000', '10010001', '10010110', '10010111',
		'10010100', '10010101', '11101010', '11101011', '11101000', '11101001', '11101110', '11101111', '11101100', '11101101',
		'11100010', '11100011', '11100000', '11100001', '11100110', '11100111', '11100100', '11100101', '11111010', '11111011',
		'11111000', '11111001', '11111110', '11111111', '11111100', '11111101', '11110010', '11110011', '11110000', '11110001' ]

		# 0, 1, ... 9
		hectometersCode = [
		'000', '001', '110', '111', '100', '101', '0110', '0111', '0100', '0101' ]

		# 1, 2, ... 10
		kilometersCode = [
		'10', '011', '001', '000', '111', '0101', '0100', '1101', '11001', '11000' ]

		# 0, 1, ... 99
		centimetersCode = [
		'110010', '110011', '110000', '110001', '110110', '110111', '110100', '110101', '0101010', '0101011',
		'0101000', '0101001', '0101110', '0101111', '0101100', '0101101', '0100010', '0100011', '0100000', '0100001',
		'0100110', '0100111', '0100100', '0100101', '0111010', '0111011', '0111000', '0111001', '0111110', '0111111',
		'0111100', '0111101', '0110010', '0110011', '0110000', '0110001', '0110110', '0110111', '0110100', '0110101',
		'0001010', '0001011', '0001000', '0001001', '0001110', '0001111', '0001100', '0001101', '0000010', '0000011',
		'0000000', '0000001', '0000110', '0000111', '0000100', '0000101', '0011010', '0011011', '0011000', '0011001',
		'0011110', '0011111', '0011100', '0011101', '0010010', '0010011', '0010000', '0010001', '0010110', '0010111',
		'0010100', '0010101', '101010', '101011', '101000', '101001', '101110', '101111', '101100', '101101',
		'100010', '100011', '100000', '100001', '100110', '100111', '100100', '100101', '1111010', '1111011',
		'1111000', '1111001', '1111110', '1111111', '1111100', '1111101', '111010', '111011', '111000', '111001' ]

		distance_mm = int(round(np.float32(distance) * np.float32(1000))) # distance in mm
		if print_cfg:
		print('distance: ', distance_mm)

		mm = distance_mm # [mm]
		cm = mm // 10 # [cm]
		m = cm // 100 # [m]
		hm = m // 100 # [hm]
		km = hm // 10 # [km]

		mm = (mm % 10) # 0, 1, ... 9 [mm]
		cm = (cm % 100) # 0, 1, ... 99 [cm]
		m = (m % 100) # 0, 1, ... 99 [m]
		hm = (hm % 9) # 0, 1, ... 9 [hm]

		assert 0 <= mm <= 9
		assert 0 <= cm <= 99
		assert 0 <= m <= 99
		assert 0 <= hm <= 9
		assert distance_mm == km * 1000000 + hm * 100000 + m * 1000 + cm * 10 + mm

		code = metersCode[m]

		if not isSmallScene:
		# large scenes
		if hm > 0 or km > 0:
		# hectometers
		code += '1' + hectometersCode[hm]
		while km > 0:
		# kilometers
		code += '1' + kilometersCode[min(km, 10)]
		km = km - 10
		code += '0'
		else:
		code += '0'

		# centimeters
		if cm > 0:
		code += '1' + centimetersCode[cm]
		else:
		code += '0'

		return code


		def get_absorption_code(absorption):
		assert 0.0 <= absorption <= 1.0

		index = round(absorption * 10.0)

		return ['110', '100', '101', '0110', '0111', '111', '0100', '0101', '0010', '0011', '000' ][index]

		# apply function to elements of list and concatenate the resulting strings
		def concatenate(function, data):
		return ''.join([function(d) for d in data])
		@@ -248,6 +333,7 @@ def concatenate(function, data):

		def test():
		# generate binary output which can be compared with the Matlab implementation output

		string = ''

		# count or index encoding
		@@ -282,7 +368,54 @@ def generate_reverb_payload_equivalent_to_rend_config_renderer_cfg():
		# based on config_renderer.cfg
		# note that because of encoding, resolution is lost and behaviour may not be bit-exact compared to .cfg file based values
		data = bitarray(
		get_count_or_index_code(1) # fgdNrGrids
		'1' # hasAcEnv
		+ get_count_or_index_code(1) # fgdNrGrids
		+ fgdMethod.Individual_Frequencies.value # fgdMethod
		+ get_count_or_index_code(31) # fgdNrBands

		+ concatenate(get_frequency_code, # fgdCenterFreq
		[ 20.0, 25.0, 31.5, 40.0, 50.0, 63.0, 80.0, 100.0, 125.0, 160.0,
		200.0, 250.0, 315.0, 400.0, 500.0, 630.0, 800.0, 1000.0, 1250.0, 1600.0,
		2000.0, 2500.0, 3150.0, 4000.0, 5000.0, 6300.0, 8000.0, 10000.0, 12500.0, 16000.0,
		20000.0 ])

		+ get_count_or_index_code(1) # AcousticEnvCount
		+ get_id_code(0) # ID
		+ get_count_or_index_code(0) # FreqGridID
		+ get_duration_code(0.1) # (input)Predelay

		+ concatenate(get_duration_code, # RT60
		[ 1.3622, 1.4486, 1.3168, 1.5787, 1.4766, 1.3954, 1.2889, 1.3462, 1.0759, 1.0401,
		1.0970, 1.0850, 1.0910, 1.0404, 1.0499, 1.0699, 1.1028, 1.1714, 1.1027, 1.0666,
		1.0550, 1.0553, 1.0521, 1.0569, 1.0421, 0.97822, 0.80487, 0.75944, 0.71945, 0.61682,
		0.60031 ])

		+ concatenate(get_dsr_code, # DSR
		[ 1.8811e-08, 2.1428e-08, 1.3972e-08, 1.51e-08, 1.287e-08, 1.8747e-08, 2.413e-08, 3.9927e-08, 8.9719e-08, 1.902e-07,
		3.702e-07, 6.1341e-07, 7.1432e-07, 6.5331e-07, 4.6094e-07, 5.4683e-07, 7.0134e-07, 6.856e-07, 7.114e-07, 6.9604e-07,
		5.2939e-07, 5.699e-07, 6.1773e-07, 5.7488e-07, 4.7748e-07, 2.7213e-07, 1.3681e-07, 1.0941e-07, 6.2001e-08, 2.8483e-08,
		2.6267e-08 ])
		+ '0' # hasEarlyReflections
		, endian='big')

		file = open('rend_config_renderer.dat', 'wb')
		data.tofile(file)
		file.close()

		def generate_reverb_payload_equivalent_to_rend_config_renderer_cfg_plus_early_reflections_no_listener_origin():
		# based on config_renderer.cfg
		# note that because of encoding, resolution is lost and behaviour may not be bit-exact compared to .cfg file based values
		data = bitarray(
		'1' # hasAcEnv
		+ get_count_or_index_code(2) # fgdNrGrids
		# frequency grid #1
		+ fgdMethod.Individual_Frequencies.value # fgdMethod
		+ get_count_or_index_code(1) # fgdNrBands

		+ concatenate(get_frequency_code, # fgdCenterFreq
		[ 10000.0 ])

		# frequency grid #2
		+ fgdMethod.Individual_Frequencies.value # fgdMethod
		+ get_count_or_index_code(31) # fgdNrBands

		@@ -294,7 +427,64 @@ def generate_reverb_payload_equivalent_to_rend_config_renderer_cfg():

		+ get_count_or_index_code(1) # AcousticEnvCount
		+ get_id_code(0) # ID
		+ get_count_or_index_code(1) # FreqGridID
		+ get_duration_code(0.1) # (input)Predelay

		+ concatenate(get_duration_code, # RT60
		[ 1.3622, 1.4486, 1.3168, 1.5787, 1.4766, 1.3954, 1.2889, 1.3462, 1.0759, 1.0401,
		1.0970, 1.0850, 1.0910, 1.0404, 1.0499, 1.0699, 1.1028, 1.1714, 1.1027, 1.0666,
		1.0550, 1.0553, 1.0521, 1.0569, 1.0421, 0.97822, 0.80487, 0.75944, 0.71945, 0.61682,
		0.60031 ])

		+ concatenate(get_dsr_code, # DSR
		[ 1.8811e-08, 2.1428e-08, 1.3972e-08, 1.51e-08, 1.287e-08, 1.8747e-08, 2.413e-08, 3.9927e-08, 8.9719e-08, 1.902e-07,
		3.702e-07, 6.1341e-07, 7.1432e-07, 6.5331e-07, 4.6094e-07, 5.4683e-07, 7.0134e-07, 6.856e-07, 7.114e-07, 6.9604e-07,
		5.2939e-07, 5.699e-07, 6.1773e-07, 5.7488e-07, 4.7748e-07, 2.7213e-07, 1.3681e-07, 1.0941e-07, 6.2001e-08, 2.8483e-08,
		2.6267e-08 ])

		+ '1' # hasEarlyReflections
		+ concatenate(get_distance_code, # room dimensions
		[ 3.0, 4.0, 2.5 ])

		+ get_count_or_index_code(0) # FreqGridID
		+ concatenate(get_absorption_code, # absorptionCode
		[ 0.8, 0.8, 0.8, 0.8, 0.2, 0.6 ])

		+ '0' # listener origin flag

		, endian='big')

		file = open('rend_config_renderer.dat', 'wb')
		data.tofile(file)
		file.close()


		def generate_reverb_payload_equivalent_to_rend_config_renderer_cfg_plus_early_reflections_listener_origin():
		# based on config_renderer.cfg
		# note that because of encoding, resolution is lost and behaviour may not be bit-exact compared to .cfg file based values
		data = bitarray(
		'1' # hasAcEnv
		+ get_count_or_index_code(2) # fgdNrGrids
		# frequency grid #1
		+ fgdMethod.Individual_Frequencies.value # fgdMethod
		+ get_count_or_index_code(1) # fgdNrBands

		+ concatenate(get_frequency_code, # fgdCenterFreq
		[ 10000.0 ])

		# frequency grid #2
		+ fgdMethod.Individual_Frequencies.value # fgdMethod
		+ get_count_or_index_code(31) # fgdNrBands

		+ concatenate(get_frequency_code, # fgdCenterFreq
		[ 20.0, 25.0, 31.5, 40.0, 50.0, 63.0, 80.0, 100.0, 125.0, 160.0,
		200.0, 250.0, 315.0, 400.0, 500.0, 630.0, 800.0, 1000.0, 1250.0, 1600.0,
		2000.0, 2500.0, 3150.0, 4000.0, 5000.0, 6300.0, 8000.0, 10000.0, 12500.0, 16000.0,
		20000.0 ])

		+ get_count_or_index_code(1) # AcousticEnvCount
		+ get_id_code(0) # ID
		+ get_count_or_index_code(1) # FreqGridID
		+ get_duration_code(0.1) # (input)Predelay

		+ concatenate(get_duration_code, # RT60
		@@ -309,6 +499,20 @@ def generate_reverb_payload_equivalent_to_rend_config_renderer_cfg():
		5.2939e-07, 5.699e-07, 6.1773e-07, 5.7488e-07, 4.7748e-07, 2.7213e-07, 1.3681e-07, 1.0941e-07, 6.2001e-08, 2.8483e-08,
		2.6267e-08 ])

		+ '1' # hasEarlyReflections
		+ concatenate(get_distance_code, # room dimensions
		[ 3.0, 4.0, 2.5 ])

		+ get_count_or_index_code(0) # FreqGridID
		+ concatenate(get_absorption_code, # absorptionCode
		[ 0.8, 0.8, 0.8, 0.8, 0.2, 0.6 ])

		+ '1' # listener origin flag
		+ '1' # positive x origin
		+ '0' # negative y origin
		+ concatenate(get_distance_code, # listener origin (x, y, z)
		[ 0.5, 0.5, 1.5 ])

		, endian='big')

		file = open('rend_config_renderer.dat', 'wb')
		@@ -316,11 +520,13 @@ def generate_reverb_payload_equivalent_to_rend_config_renderer_cfg():
		file.close()



		def generate_reverb_payload_equivalent_to_rend_config_hospital_patientroom_cfg():
		# based on config_hospital_patientroom.cfg
		# note that because of encoding, resolution is lost and behaviour may not be bit-exact compared to .cfg file based values
		data = bitarray(
		get_count_or_index_code(1) # fgdNrGrids
		1 # hasAcEnv
		+ get_count_or_index_code(1) # fgdNrGrids
		+ fgdMethod.Individual_Frequencies.value # fgdMethod
		+ get_count_or_index_code(31) # fgdNrBands

		@@ -348,6 +554,8 @@ def generate_reverb_payload_equivalent_to_rend_config_hospital_patientroom_cfg()
		6.379496e-06, 6.358105e-06, 6.6696e-06, 6.369334e-06, 6.378474e-06, 3.339913e-06, 3.129318e-06, 2.892564e-06, 6.00202e-07, 3.40124e-07,
		3.37705e-07 ])

		+ '0' # hasEarlyReflections

		, endian='big')

		file = open('rend_config_hospital_patientroom.dat', 'wb')
		@@ -359,7 +567,8 @@ def generate_reverb_payload_equivalent_to_rend_config_recreation_cfg():
		# based on config_recreation.cfg
		# note that because of encoding, resolution is lost and behaviour may not be bit-exact compared to .cfg file based values
		data = bitarray(
		get_count_or_index_code(1) # fgdNrGrids
		1 # hasAcEnv
		+ get_count_or_index_code(1) # fgdNrGrids
		+ fgdMethod.Individual_Frequencies.value # fgdMethod
		+ get_count_or_index_code(31) # fgdNrBands

		@@ -387,6 +596,8 @@ def generate_reverb_payload_equivalent_to_rend_config_recreation_cfg():
		2.958383e-06, 2.725911e-06, 7.94912e-07, 6.20198e-07, 5.71181e-07, 5.5546e-08, 1.3987e-08, 1.338e-08, 1.322e-09, 1.3e-11,
		4e-12 ])

		+ '0' # hasEarlyReflections

		, endian='big')

		file = open('rend_config_recreation.dat', 'wb')