Date: July, 2023
Based on the Paper:
A Novel Approach of Decoding EEG Four-class Motor Imagery Tasks via Scout ESI and CNN
The method in this repository is EEG source imaging (ESI) + Fourier Transform for joint time-frequency analysis + Convolutional Neural Networks (CNNs). The raw data has been processed using the Matlab Toolkit Brainstorm. The technique of ESI uses a boundary element method (BEM) and weighted minimum norm estimation (WMNE) to solve the EEG forward and inverse problems, respectively. Ten scouts are then created within the motor cortex to select the region of interest (ROI). Features were extracted from the time series of scouts using a Morlet wavelet approach. Lastly, CNN is employed for classifying MI tasks.
Layer (type) Output Shape Param #
Conv2d-1 [-1, 32, 32, 20] 320
Dropout-2 [-1, 32, 32, 20] 0
Conv2d-3 [-1, 32, 32, 20] 9,248
BatchNorm2d-4 [-1, 32, 32, 20] 64
Conv2d-5 [-1, 64, 32, 20] 36,928
Dropout-6 [-1, 64, 32, 20] 0
Conv2d-7 [-1, 64, 14, 8] 36,928
BatchNorm2d-8 [-1, 64, 14, 8] 128
Dropout-9 [-1, 64, 14, 8] 0
Conv2d-10 [-1, 64, 14, 8] 36,928
BatchNorm2d-11 [-1, 64, 14, 8] 128
Conv2d-12 [-1, 128, 14, 8] 147,584
Dropout-13 [-1, 128, 14, 8] 0
Flatten-14 [-1, 3584] 0
Linear-15 [-1, 512] 1,835,520
BatchNorm2d-16 [-1, 512, 1, 1] 1,024
Dropout-17 [-1, 512] 0
Linear-18 [-1, 4] 2,052
Total params: 2,106,852
Trainable params: 2,106,852
Non-trainable params: 0
