|
|
---
|
|
|
tags:
|
|
|
- EEG
|
|
|
- EEGNet
|
|
|
- biometric-identification
|
|
|
- brain-computer-interface
|
|
|
- biosignals
|
|
|
- neuroscience
|
|
|
- deep-learning
|
|
|
- tensorflow
|
|
|
- physiological-biometrics
|
|
|
- eegmmidb
|
|
|
- neurotechnology
|
|
|
- brain-signals
|
|
|
---
|
|
|
# EEGNet Identification Model |
|
|
|
|
|
## Model Description |
|
|
A deep learning model for EEG-based subject identification using the EEG Motor Movement/Imagery Dataset. Based on the EEGNet architecture implemented in TensorFlow/Keras. |
|
|
|
|
|
- **Input**: 3-second EEG segments (64 channels, 160 Hz sampling rate) |
|
|
- **Output**: Predicted subject ID (109 possible subjects) |
|
|
- **Architecture**: Compact CNN with: |
|
|
- Temporal/spatial convolutions |
|
|
- Batch normalization |
|
|
- Dropout regularization |
|
|
- Softmax classification layer |
|
|
|
|
|
## Dataset |
|
|
**EEG Motor Movement/Imagery Dataset (EEGMMIDB)** |
|
|
109 subjects, 64 channels (10-20 system) |
|
|
|
|
|
**Preprocessing Pipeline**: |
|
|
1. Bandpass filtering (1–40 Hz) |
|
|
2. Resampling to 160 Hz |
|
|
3. Segmentation into 3-second non-overlapping windows |
|
|
4. Channel name normalization |
|
|
5. Standard 10-20 montage alignment |
|
|
|
|
|
## Training |
|
|
| Parameter | Value | |
|
|
|------------------------|---------------------------| |
|
|
| **Loss Function** | Categorical Crossentropy | |
|
|
| **Optimizer** | Adam (lr=0.001) | |
|
|
| **Batch Size** | 64 | |
|
|
| **Max Epochs** | 100 (early stopping) | |
|
|
| **Validation Split** | 15% | |
|
|
| **Test Split** | 15% | |
|
|
|
|
|
**Callbacks**: |
|
|
- EarlyStopping (patience=8) |
|
|
- ModelCheckpoint (best weights) |
|
|
- ReduceLROnPlateau (factor=0.2, patience=5) |
|
|
- TensorBoard logging |
|
|
|
|
|
## Evaluation |
|
|
**Primary Metric**: Classification Accuracy |
|
|
**Test Performance**: 93.97% |
|
|
|
|
|
**Analysis Tools**: |
|
|
- Overall/per-subject accuracy |
|
|
- Confusion matrix (top 15 subjects) |
|
|
- Classification report (precision/recall/F1 per subject) |
|
|
- Training/validation curves |
|
|
- Filter visualizations (temporal + spatial) |
|
|
- Model attention maps for sample segments |
|
|
|
|
|
## Usage |
|
|
### Inference Requirements |
|
|
**Input Shape**: `(64, 480)` (channels × time points) |
|
|
**Critical**: Input must match preprocessing: |
|
|
- Bandpass filtered (1-40 Hz) |
|
|
- Resampled to 160 Hz |
|
|
- Standard 10-20 montage channels |
|
|
- Identical normalization as training data |
|
|
|
|
|
### Visualization |
|
|
Code templates available for: |
|
|
1. Training history plots |
|
|
2. Confusion matrix generation |
|
|
3. Spatial filter visualization |
|
|
4. Attention heatmaps for EEG segments |
|
|
|
|
|
## Limitations |
|
|
1. Trained/tested exclusively on EEGMMIDB |
|
|
2. Performance not guaranteed for: |
|
|
- Different EEG systems |
|
|
- Alternative montages |
|
|
- Varying preprocessing pipelines |
|
|
3. Requires strict adherence to original preprocessing for reliable inference |
|
|
|
|
|
## Citation |
|
|
```bibtex |
|
|
@article{lawhern2018eegnet, |
|
|
title={EEGNet: A Compact Convolutional Neural Network for EEG-based Brain–Computer Interfaces}, |
|
|
author={Lawhern, Vernon J. and Solon, Amelia J. and Waytowich, Nicholas R. and Gordon, Stephen M. and Hung, Chou P. and Lance, Brent J.}, |
|
|
journal={Journal of Neural Engineering}, |
|
|
volume={15}, |
|
|
number={5}, |
|
|
pages={056013}, |
|
|
year={2018}, |
|
|
publisher={IOP Publishing} |
|
|
} |
|
|
|
|
|
@article{goldberger2000physiobank, |
|
|
title={PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals}, |
|
|
author={Goldberger, Ary L. and Amaral, Luis A. N. and Glass, Leon and Hausdorff, Jeffrey M. and Ivanov, Plamen Ch. and Mark, Roger G. and Mietus, Joseph E. and Moody, George B. and Peng, Chung-Kang and Stanley, H. Eugene}, |
|
|
journal={Circulation}, |
|
|
volume={101}, |
|
|
number={23}, |
|
|
pages={e215--e220}, |
|
|
year={2000} |
|
|
} |
|
|
``` |
