EEGProc documentation

Contents:

• Getting Started
  • Installation
  • Dependencies
  • Quick Start
  • Documentation Structure
• Convert public datasets to CSV
  • Get the datasets
  • Common behavior
  • AMIGOS
  • DREAMER
  • EEGEmotions-27
  • Cowen 27 emotion mapping
  • DEAP
  • Use the result
• eegproc
  • eegproc package

EEGProc is a vectorized library for preprocessing EEG (electroencephalogram) data, extracting features from it, and evaluating models on it subject by subject. It is built for researchers and developers working in neuroscience, biomedical engineering, and machine learning. Check out and star or fork the project at https://github.com/EEGo-UNC/EEGProc

Features

• Preprocessing: detrending, interpolation, notch and band-pass filtering into frequency bands.

• Featurization: band power, Shannon, wavelet and IMF energy and entropy, and Hjorth parameters, computed on pandas DataFrames.

• Dataset conversion: eegproc-to-csv turns downloaded AMIGOS, DREAMER, EEGEmotions-27 and DEAP files into tidy CSV tables.

• Subject-wise evaluation (eegproc[deep-learning]): leave-one-subject-out, nested and few-shot calibration cross-validation straight from a tidy table.

• Model components (eegproc[deep-learning]): CNN, GNN and RNN encoders, variational classifier heads, contrastive and autoencoder losses, and domain-generalization helpers.

• Counterfactual explanations (eegproc[deep-learning]): adapter-based counterfactual optimization for differentiable models, with scalp topographies.

Contributing

Contributions are welcome! If you have ideas for new features or improvements, feel free to open an issue or submit a pull request.

License

This project is licensed under the GPLv2 License.