🧠 Brain Activity Signal Processing (EEG/MEG)

A comprehensive Python toolbox for processing EEG/MEG recordings, extracting band‑specific power features, designing digital filters, visualizing spectrograms, and building a motor‑imagery Brain–Computer Interface (BCI) application.

---

📂 Project Structure

Brain_Activity_Signal_Processing_EEG_MEG/
│
├── dataset/                           # (Raw data - not tracked by git)
│   └── files/
│       └── S001/
│           ├── S001R01.edf
│           └── S001R01.edf.event
│
├── src/                               # All source code
│   ├── __init__.py                    # Package initialization
│   ├── load_data.py                   # EDF loading, filtering, event parsing
│   ├── extract_band_specific_power.py # Framing & DFT-based PSD extraction
│   ├── design_filters.py              # Frequency response & pole-zero plots
│   ├── plot_spectrograms.py           # Spectrogram & dominant rhythm plots
│   └── application_bci.py             # Motor-imagery BCI (LDA classifier)
│
├── results/                           # (Optional) Saved models and plots
│
├── requirements.txt                   # Python dependencies
└── README.md                          # This file

---

⚙️ Installation & Dependencies

1. Clone the Repository

git clone https://github.com/mahajialirezaei/Brain_Activity_Signal_Processing_EEG_MEG.git
cd Brain_Activity_Signal_Processing_EEG_MEG

2. (Optional) Create a Virtual Environment

python -m venv venv
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activate

3. Install Required Packages

pip install -r requirements.txt

Contents of requirements.txt (if you don't have it yet):

mne
numpy
scipy
matplotlib
scikit-learn
joblib

---

🚀 How to Run

Note: All scripts are now inside the src/ folder. Run them from the project root using the python -m src.<module> syntax, or directly with python src/<file>.py.

1. Process & Load Data (Basic Inspection)

python -m src.load_data

or

python src/load_data.py

• Reads the EDF files defined in load_data.py.
• Applies bandpass filters (α: 8–12Hz, β: 13–30Hz, γ: 30–45Hz).
• Computes absolute band powers using Welch's method.
• Displays a spectrogram for the first channel.

2. Extract Band-Specific Power Dynamics

python -m src.extract_band_specific_power

• Frames the signal into overlapping windows.
• Computes PSD per frame using DFT.
• Generates plots for average PSD and power-time series per band.

3. Visualize Filter Designs

python -m src.design_filters

• Plots Magnitude/Phase response and Pole-Zero diagrams for the α, β, and γ Butterworth bandpass filters.

4. Run the Motor-Imagery BCI Application

python -m src.application_bci

• Loads specific runs (R04, R08, R12) for motor imagery.
• Extracts band power features from channels C3.. and C4...
• Trains an LDA (Linear Discriminant Analysis) classifier.
• Prints test accuracy, 5-fold cross-validation scores.
• Displays an ROC curve.
• Simulates a real-time online classification (LEFT vs. RIGHT imagery) using a sliding window.

---

⚙️ Configuration

All configurable variables are centralized in src/load_data.py and src/application_bci.py:

File	Variable	Description
load_data.py	subjects	List of subjects (e.g., ['S001', 'S002'])
load_data.py	runs	List of runs (e.g., ['R01', 'R02', ...])
load_data.py	BANDS	Frequency bands (alpha, beta, gamma)
application_bci.py	imagery_runs	Runs used for motor imagery (default: R04, R08, R12)
application_bci.py	target_channels	Channels for feature extraction (default: C3.., C4..)
application_bci.py	window_sec	Window length in seconds for feature extraction

💡 Path Handling: After restructuring, load_data.py automatically resolves the dataset path using os.path.dirname(os.path.dirname(__file__)). It assumes the dataset/ folder is located in the project root.

---

📖 Module Descriptions

src/load_data.py

• Core Functions:
  • process_edf_file(file_path): Reads EDF, applies filters, extracts events, computes band powers.
  • getSubject(), getRun(), getbase_dir(), getBands(): Utility getters.

src/extract_band_specific_power.py

• Core Logic:
  • frame_signal(): Overlapping sliding windows.
  • compute_psd_dft(): Periodogram via FFT.
  • band_power_from_psd(): Integrates PSD over specific frequency bands.
  • Calls plot_spectrogram_and_dominant() from plot_spectrograms.py.

src/design_filters.py

• Core Logic:
  • Retrieves filter coefficients from load_data.py.
  • Plots magnitude/phase response and pole-zero maps for each filter.

src/plot_spectrograms.py

• Core Function:
  • plot_spectrogram_and_dominant(): Displays a spectrogram (0–50Hz) and a scatter plot of the dominant rhythm over time.

src/application_bci.py

• Core Functions:
  • extract_motor_imagery_features(): Builds feature matrix (X) and labels (y) for imagery events.
  • train_lda(X, y): Splits data, trains an LDA, evaluates with accuracy, CV, and ROC.
  • simulate_online(clf): Simulates real-time decoding with overlapping windows.

---

🧪 Example Output (BCI Application)

After running python -m src.application_bci, you should see something like:

Test accuracy: 0.78
5-fold CV: 0.75 ± 0.04
Simulated online motor imagery classification:
0.00-2.00s -> LEFT
2.00-4.00s -> RIGHT
...

An ROC curve plot will also appear, showing the AUC score.

---

📚 References

• MNE-Python Documentation
• Welch's method for Power Spectral Density estimation.
• Linear Discriminant Analysis for feature classification in BCI.
• Dataset: PhysioNet EEG Motor Imagery Dataset.

---

👨‍💻 Developer

Developed by Mohammad Amin Haji Alirezaei

Feel free to ⭐️ this repository, open an issue, or contribute!

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.