Table 1. The detailed parameters of our proposed deep neural network.

Layers	Layer name	Kernel size	No. of filters	Stride	Output shape	No. of trainable parameters	No. of non-trainable parameters
0	Input1a	–	–	–	200 × 1	–	–
1	1D Convolution	11	16	3	64 × 16	192	–
2	Batch Normalization	–	–	–	64 × 16	32	32
3	ReLU	–	–	–	64 × 16	–	–
4	Max-Pooling	3	–	2	31 × 16	–	–
5	1D Convolution	5	32	1	27 × 32	2,592	–
6	Batch Normalization	–	–	–	27 × 32	64	64
7	ReLU	–	–	–	27 × 32	–	–
8	Max-Pooling	3	–	2	13 × 32	–	–
9	1D Convolution	3	64	1	11 × 64	6,208	–
10	Batch Normalization	–	–	–	11 × 64	128	128
11	ReLU	–	–	–	11 × 64	–	–
12	Max-Pooling	3	–	2	5 × 64	–	–
13	Flatten	–	–	–	320	–	–
14	Input2b	–	–	–	4	–	–
15	Concatenate	–	–	–	324	–	–
16	Dense	–	–	–	64	20,800	–
17	Dense	–	–	–	4	260	–

Notes:

^aRefers to the raw signal of the heartbeat. The morphological features of the heartbeat will be obtained through the CNN architecture.

^bIs the RR interval features of the heartbeat. It will be combined with the CNN-based morphological features to build the final classification model.