Automatic Detection of Obstructive Sleep Apnea Events Using a Deep CNN-LSTM Model

Table 3

The parameters of the proposed model.

Layer	Layer type	Units	Unit type	Size	Stride	Output size
Input						1000 × 1
BN						1000 × 1
Cn_1	Convolutional	24	ReLU	125 × 1	1 × 1	876 × 24
Cn_2	Convolutional	24	ReLU	15 × 1	1 × 1	986 × 24
Cn_3	Convolutional	24	ReLU	5 × 1	1 × 1	996 × 24
Mp_1	Max pooling	24		2 × 1	1 × 1	438 × 24
Mp_2	Max pooling	24		2 × 1	1 × 1	493 × 24
Mp_3	Max pooling	24		2 × 1	1 × 1	498 × 24
Concatenate		24				1429 × 24
Mp_4	Max pooling	24		3 × 1	1 × 1	476 × 24
Add	Add	24				1000 × 24
Dense	Fully connected	48	LeakyReLU			1000 × 48
Dropout	Dropout					1000 × 48
Gp	Global pooling					48 × 1
LSTM	LSTM					64 × 1
Dense	Fully connected	2	Softmax			2