Table 1 Layers of semantic generation module and output dimensions of each layer

Convolutional part

Input

2 × 1024 × 1

Conv1D (filters 64, size 1 × 5)

64 × 1016 × 1

Maxpooling (size 1 × 5)

64 × 204 × 1

Conv1D (filters 64, size 1 × 5)

64 × 200 × 1

Maxpooling (size 1 × 5)

64 × 40 × 1

Conv1D (filters 64, size 1 × 5)

64 × 36 × 1

Maxpooling (size 1 × 5)

64 × 8 × 1

Conv1D (filters 64, size 1 × 5)

64 × 4 × 1

Maxpooling (size 1 × 5)

64 × 1 × 1

Flatten

64 × 1

Fully connected

64 × 1

Fully connected

64 × 1

Encoding part

Encoding

32 × 47 × 1

Conv1D (filters 128, size 1 × 5)

128 × 43 × 1

Maxpooling (size 1 × 5)

128 × 9 × 1

ReLU

–

Conv1D (filters 128, size 1 × 5)

128 × 5 × 1

Maxpooling (size 1 × 5)

128 × 1 × 1

ReLU

–

Flatten

128 × 1

Fully connected

64 × 1