This repository provides a framework for extracting universal, task-independent semantic features from EEG signals. It addresses the limitations of traditional task-specific EEG feature extraction methods by integrating CNNs, AutoEncoders, and Transformers. The resulting high-level semantic representations are robust to inter-subject variability and applicable across diverse EEG paradigms.
We plan to evolve this repository into a more user-friendly, fully automated framework. Users will specify dataset parameters (e.g., sampling rate, channels), and the system will handle preprocessing and feature extraction, streamlining reproducibility and accessibility for the EEG research community.