Research Article
A Novel Deep Neural Network for Robust Detection of Seizures Using EEG Signals
Table 1
Details of the CNN structure used in this research.
| Block | Type | Number of neurons | Kernel size for each | Stride | (Output layer) | Output feature map |
| Conv 1 | Convolution | 139 × 20 | 40 | 1 | BN | 139 × 20 | — | — | ReLU | 139 × 20 | — | — | Dropout | 139 × 20 | — | — | Max-pooling | 70 × 20 | 2 | 2 |
| Conv 2 | Convolution | 51 × 40 | 20 | 1 | BN | 51 × 40 | — | — | ReLU | 51 × 40 | — | — | Dropout | 51 × 40 | — | — | Max-pooling | 26 × 40 | 2 | 2 |
| Conv 3 | Convolution | 17 × 80 | 10 | 1 | BN | 17 × 80 | — | — | ReLU | 17 × 80 | — | — | Dropout | 17 × 80 | — | — | Max-pooling | 9 × 80 | 2 | 2 |
| FC 1 | FC | 64 | — | — | ReLU | 64 | — | — | Dropout | 64 | — | — |
| FC 2 | FC | 32 | — | — | ReLU | 32 | — | — | Dropout | 32 | — | — | FC 3 | FC | 2 or 3 or 5 | — | — |
|
|