Documentation

Setup, usage guide, and API reference for the BCI Classifier system.

Getting Started

Installation

# Clone repository
git clone https://github.com/Shaheer2492/BCI-Classifer.git
cd BCI-Classifer

# Create environment
python -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

Key Scripts

src/generate_ground_truth_labels.py Phase 1: Processes PhysionetMI data to calculate baseline decoder accuracy.

src/extract_early_trial_features.py Phase 2: Extracts spectral and spatial features from the first 15 trials.

src/train_performance_predictor.py Phase 3: Trains ML models (Random Forest) to predict final accuracy.

src/prediction_server.py Phase 4: Flask API that serves the trained model for the web interface.

API Reference

PhysionetMIGroundTruthGenerator

Main class for generating ground truth labels.

n_subjects (int): Number of subjects (default 109).
generate_all_labels(output_path): Runs the pipeline.

EarlyTrialFeatureExtractor

Extracts features from subsets of data.

n_early_trials (int): Number of trials to use for prediction.
extract_subject_features(subject_id): Returns 23-dim feature vector.

Troubleshooting

Dataset Download: Validates internet connection for 2.5GB Physionet download.
Memory: Processing all 109 subjects usually takes 1-2 hours.
Viz: Ensure prediction_server.py is running before opening the Interface.