What problem does it solve? Running an end-to-end analysis on the SEED-IV emotion EEG dataset requires coordinating BIDS validation, EEG preprocessing, feature extraction, and classification across multiple tools, which is error-prone when done manually. ## Core Features & Use Cases - BIDS Validation: Checks SEED-IV directory structure, subject/session completeness (15 subjects x 3 sessions), and EEG/event file presence, producing a compliance report. - Feature Extraction: Computes Differential Entropy (DE) and Power Spectral Density (PSD) features across delta, theta, alpha, beta, and gamma bands from 62-channel EEG. - Emotion Classification: Trains SVM and Random Forest classifiers with Leave-One-Subject-Out cross-validation for 4-class emotion recognition (happy, sad, fear, neutral). - Use Case: A researcher downloads SEED-IV data, asks for the full pipeline, and receives a numbered execution plan; after confirmation, the skill validates the BIDS layout, delegates preprocessing to eeg-skill, extracts DE features, and outputs classification accuracies into a clean seed_iv_output/ directory. ## Quick Start Ask the agent to run the full SEED-IV pipeline on your local SEED-IV BIDS directory, including validation, DE feature extraction, and SVM emotion classification.
