Table 1. CNN-based architecture details.

The details of the architecture of CNN_3, CNN_4, and CNN_LSTM are described. Both CNN_3 and CNN_4 are CNN-based architecture, but they are different in the number of layers and the filter in each layer. The CNN_LSTM is a hybrid CNN with bi-directional LSTM.

Name	Architectures
	Layers	Details
CNN_3	conv2D layer 1	70 filters of size (9,4)
	dropout layer 1	p = 0.2
	conv2D layer 2	100 filters of size (7,1)
	maxpool layer 1	pool size (2,1)
	dropout layer 2	p = 0.2
	conv2D layer 3	150 filters of size (7,1)
	maxpool layer 2	pool size (2,1)
	dropout layer 3	p = 0.2
	dense layer 1	512 neurons
	dropout layer 4	p = 0.2
	softmax layer	2 outputs
CNN_4	conv2D layer 1	70 filters of size (3,4)
	dropout layer 1	p = 0.2
	conv2D layer 2	100 filters of size (3,1)
	dropout layer 2	p = 0.2
	conv2D layer 3	100 filters of size (3,1)
	maxpool layer 1	pool size (2,1)
	dropout layer 3	p = 0.2
	conv2D layer 4	200 filters of size (3,1)
	maxpool layer 2	pool size (2,1)
	dropout layer 4	p = 0.2
	dense layer 1	512 neurons
	dropout layer 5	p = 0.2
	softmax layer	2 outputs
CNN_LSTM	conv1D layer 1	320 filters of length 26
	maxpool layer 1	pool size (13)
	dropout layer 1	p = 0.2
	bidirectional LSTM layer 1	320 output dimension
	dropout layer 2	p = 0.5
	Dense layer 1	925 neurons
	softmax layer	2 outputs