Project Overview
SyncTew converts EEG signals into live controls for a racing game. Using affordable biosensing hardware, the system classifies whether the player is focused or relaxed and maps that to accelerate and brake—all in real time.
Why BCIs?
BCIs translate brain activity into device commands. Beyond gaming, the same workflow can unlock assistive interfaces, neurofeedback, and research tools. SyncTew explores how far you can go with off-the-shelf components and disciplined signal processing.
Technical Implementation
Hardware
- BioAmp EXG Pill + Cable v3
- Gel electrodes + skin prep gel (frontal placement)
- Arduino Uno / Maker Uno (512 Hz sampling)
Software & pipeline
- Arduino IDE streams EEG to Python.
- Python stack: filtering, feature extraction (bandpower, variance), SVM classifier.
- Web interface receives classified state and drives the game (attentive → accelerate, relaxed → brake).
- Capture: electrodes on prefrontal cortex feed BioAmp → Arduino.
- Preprocess: notch + band-pass filtering removes noise and drift.
- Extract: compute time/frequency features in sliding windows.
- Classify: SVM predicts state (attentive vs relaxed).
- Control: mapped to accelerate/brake events in the racing game.
Research Methodology
Data Collection
- EEG data was collected in two distinct mental states: attentive and relaxed
- 20 minutes of data was recorded for each state
- Sampling rate: 512 Hz (512 data points per second)
Model Training
- Features were extracted from the raw EEG signals
- A Support Vector Machine (SVM) classifier was trained on the labeled data
- Model achieved approximately 85% accuracy in distinguishing between the two mental states
Results & Impact
- Real-time classification latency ≈120 ms after filtering.
- ~85% offline accuracy distinguishing attentive vs relaxed.
- Gameplay sessions feel responsive and intuitive—no button presses required.
Future Directions
- Classify additional mental states for richer control schemes.
- Port the hardware stack to a wireless/portable form factor.
- Collect larger datasets to train more robust ML models.
- Prototype accessibility use cases beyond gaming.
Acknowledgments
This project was developed as part of my B.Tech Computer Science program at VIT Bhopal University.