Research Article

CNN-Based Personal Identification System Using Resting State Electroencephalography

Table 1

Neural network parameters.

LayerOutput shapeDescription

Input(None, 1, P, C)
ICA(None, 1, P, 64)Linear, in_channels: C, out_channels:64
Conv_1(None, 32, P/2, 64)Conv2d, in_channels: 1, out_channels:32, kernel: 5 × 3, stride: (2, 1), padding: (2, 1), activation: ELU
Pool_1(None, 32, P/4, 64)MaxPool2d, kernel: 2 × 1, stride: (2, 1)
Conv_2(None, 32, P/4, 64)Conv2d, in_channels: 32, out_channels: 32, kernel: 3 × 3, stride: (1, 1), padding: (1, 1), activation: ELU
Pool_2(None, 32, P/4, 32)MaxPool2d, kernel: 1 × 2, stride: (1, 2)
Conv_3(None, 32, P/4, 32)Conv2d, in_channels: 32, out_channels: 32, kernel: 3 × 3, stride: (1, 1), padding: (1, 1), activation: ELU
Pool_3(None, 32, P/8, 32)MaxPool2d, kernel: 2 × 1, stride: (2, 1)
Flatten(None, 32×P/8×32)Flatten
FC_1(None, 512)Linear, in_channels: 32×P/8×32, out_channels: 512
Dropout(None, 512)Dropout, p: 0.5
FC_2(None, O)Linear, in_channels: 512, out_channels: O
Softmax(None, O)log_softmax