A Cascaded Convolutional Neural Network for Assessing Signal Quality of Dynamic ECG

Table 3

The configuration of CNN1 and CNN2 in the dynamic ECG signal quality assessment model.

CNN1	CNN2
Input (257 × 63)	Input (1 × 512)
↓	↓
N1_block1_32	N1_block2_32
↓	↓
Max_pooling_4 × 2	Max_pooling_1 × 4
↓	↓
N2_block1_64	N2_block2_64
↓	↓
Max_pooling_4 × 2	Max_pooling_1 × 4
↓	↓
N3_block1_128	N3_block2_128
↓	↓
Max_pooling_4 × 4	Max_pooling_1 × 4
↓	↓
Conv_1 × 1_32	Conv_1 × 1_32
Fc_128 (dropout)
Softmax

The vertical arrows indicate sequential connections in the network. Note that N1, N2, and N3 are the number of blocks. N1_block1_32 indicates that N1 block1 items were connected sequentially, each of which outputs 32 feature maps. Max_pooling_4 × 2 denotes a max pooling with a size of 4 × 2.