This project explores EEG-based signal classification for motor imagery tasks using Convolutional Neural Networks (CNNs). The goal is to develop a Brain-Computer Interface (BCI) system capable of decoding neural signals associated with motor imagery, paving the way for applications in assistive technology, neurorehabilitation, gaming, and human-computer interaction.
We use the EEG Motor Movement/Imagery Dataset from PhysioNet:
- EEGMMIDB Dataset
- 109 subjects, 64-channel EEG setup
- Imagery movement tasks: left hand, right hand, both hands, both feet
- Sampling rate: 160 Hz
- Load EEG recordings and segment them into four classes: left, right, forward, backward
- Filter noise and artifacts
- Normalize signal amplitudes
- Fast Fourier Transform (FFT): Extract frequency-domain features
- Discrete Wavelet Transform (DWT): Capture time-frequency characteristics
- CNN Feature Learning: Extract spatial-temporal patterns from EEG signals
- Neural Network Architectures: CNNs, Hybrid CNN-ANN models
- Hyperparameter Tuning: Grid search, dropout, batch normalization
- Evaluation Metrics: Accuracy, Precision, Recall, F1-score
- Cross-validation: Assess generalization performance
- Model comparisons based on evaluation metrics
- Performance visualization using confusion matrices and ROC curves
- Findings interpretation for motor imagery classification
- Implement real-time EEG signal processing
- Optimize model performance for low-latency applications
- Extend to multi-class motor imagery tasks
- Roohi-Azizi, M., et al. (2017). Changes of the brain's bioelectrical activity in cognition, consciousness, and some mental disorders. Medical Journal of the Islamic Republic of Iran, 31, 53.
- Wang, R. (2021). 5 Basics of EEG 101: Data Collection, Processing & Analysis. iMotions Blog.
- Schalk, G., et al. (2004). BCI2000: A General-Purpose Brain-Computer Interface System. IEEE Transactions on Biomedical Engineering.
- Goldberger, A., et al. (2000). PhysioBank, PhysioToolkit, and PhysioNet. Circulation.