Table 6.

Description of the 1D-CNN-based model.

	Input Layer:	1D grayscale image			[1, 7710, 1]
Feature extraction: three blocks of convolutional-normalization-activation-pooling layers
	Layers	Hyperparameters			Feature maps
Block 1	Convolutional	filters: 32	filter size: [1, 9]	stride size: [1, 1]	32 × [1, 7710]
	Batch normalization				32 × [1, 7710]
	Activation	ReLU			32 × [1, 7710]
	Pooling	max	filter size: [1, 6]	stride size: [1, 6]	32 × [1, 1285]
Block 2	Convolution	filters: 32	filter size: [1, 9]	stride size: [1, 1]	32 × [1, 1285]
	Batch normalization				32 × [1, 1285]
	Activation	ReLU			32 × [1, 1285]
	Pooling	max	filter size: [1, 6]	stride size: [1, 6]	32 × [1, 215]
Block 3	Convolutional	filters: 32	filter size: [1, 9]	stride size: [1, 1]	32 × [1, 215]
	Batch normalization				32 × [1, 215]
	Activation	ReLU			32 × [1, 215]
	Pooling	max	filter size: [1, 6]	stride size: [1, 6]	32 × [1, 36]
	Flatten layer				[1152]
4—class classification	Fully connected layer				[256]
	Fully connected layer				[128]
	Fully connected layer				[4]
	Softmax layer				[4]
	Classoutput layer	Cross-entropy loss function			[4]