Table 3.

Architecture of different Conv1D CNN model variants for the signal and derivative inputs. The output shape comprises (the number of nodes and the number of channels). The Conv1D layer parameters can be set as trainable or non-trainable in different model variants.

Layer	# of Classes = 3			# of Classes = 4			# of Classes = 5
	Output Shape	${\mathit{N}}_{\mathit{K}}=32$	${\mathit{N}}_{\mathit{K}}=64$	Output Shape	${\mathit{N}}_{\mathit{K}}=32$	${\mathit{N}}_{\mathit{K}}=64$	Output Shape	${\mathit{N}}_{\mathit{K}}=32$	${\mathit{N}}_{\mathit{K}}=64$
		Params	Params		Params	Params		Params	Params
Signal Input Layer	(64, 1)	0	0	(64, 1)	0	0	(64, 1)	0	0
Conv1D Layer	(64, 11)	363	715	(64, 12)	396	780	(64, 13)	429	845
BatchNormalization	(64, 11)	44	44	(64, 12)	48	48	(64, 13)	52	52
Activation (Tanh)	(64, 11)	0	0	(64, 12)	0	0	(64, 13)	0	0
GlobalMaxPooling	(11)	0	0	(12)	0	0	(13)	0	0
Interval Input Layer	(4, 1)	0	0	(4, 1)	0	0	(4, 1)	0	0
Dense Layer 1 (Relu)	(4, 32)	64	64	(4, 32)	64	64	(4, 32)	64	64
Dense Layer 2 (Relu)	(4, 16)	528	528	(4, 16)	528	528	(4, 16)	528	528
Dense Layer 3 (Relu)	(4, 8)	136	136	(4, 8)	136	136	(4, 8)	136	136
Flatten	(32)	0	0	(32)	0	0	(32)	0	0
Concatenate	(43)	0	0	(44)	0	0	(45)	0	0
Softmax Output Layer	(3)	132	132	(4)	180	180	(5)	230	230
Total params		1267	1619		1352	1736		1439	1855
Trainable params		882	882		932	932		984	984
Non-trainable params		385	737		420	804		455	871