Table 5 Presents a comparison of different network model structures.
Model | Details | Input | Convolution layers | Kernel size | LSTM layers | Bi-LSTM layers | GRU | Dense | Parameter | |
|---|---|---|---|---|---|---|---|---|---|---|
CNN2 | CNN | I/Q | 4 | 8 × 2,8 × 2,8 × 2,8 × 2 | 0 | 0 | 0 | 2 | 858,123 | |
CLDNN2 | CNN+ LSTM | I/Q | 4 | 3 × 1,3 × 2,3 × 1,3 × 1 | 1 | 0 | 0 | 2 | 517,443 | |
ResNET | ResNET | I/Q | 4 | 3 × 1,3 × 2,3 × 1,3 × 1 | 0 | 0 | 0 | 2 | 3,098,283 | |
LSTM | LSTM | I/Q | 0 | 0 | 1 | 0 | 0 | 1 | 200,075 | |
PET-CGDNN | CNN + GRU + DNN | I、Q and I/Q | 2 | 2 × 8,1 × 5 | 0 | 0 | 1 | 1 | 71,871 | |
MCLDNN | CNN + LSTM | I、Q and I/Q | 5 | 8 × 2,7,7,8 × 1,5 × 2 | 2 | 0 | 0 | 3 | 405,175 | |
CC-MSNet | CNN+ LSTM+ Bi-LSTM | I、Q and I/Q | I(3 Layers) Q(3 Layers) I/Q(3 Layers) I/Q(1 Layers) | 2,4,8 2,4,8 (2,1),(1,3),(1,3) Complex-Convolution(2,3) | (9,5) | 1 | 1 | 0 | 2 | 654,687 |
