Research Article
Fusion of Deep Features from 2D-DOST of fNIRS Signals for Subject-Independent Classification of Motor Execution Tasks
Table 5
The structure of modified LeNet used in joint fusion.
| Operation | Layer | Input | Filter size | Output |
| Feature extraction | Convolution | 128 × 16 × 1 | 5 × 5 × 10 | 124 × 12 × 10 | Average pooling | 124 × 12 × 10 | Pool 2 × 2 | 62 × 6 × 10 | Convolution | 62 × 6 × 10 | 3 × 3 × 10 | 60 × 4 × 10 | Average pooling | 60 × 4 × 10 | Pool 2 × 2 | 30 × 2 × 10 | Flatten | 30 × 2 × 10 | — | 600 × 1 |
| Classification | Fully connected | 2400 × 1 | — | 128 × 1 | Fully connected | 128 × 1 | — | 64 × 1 | Fully connected | 64 × 1 | — | × 1 |
|
|