EEG-Based Mental State Classification for Real-Time Game Control Using a Low-Cost BCI System

This project presents a low-cost, real-time Brain-Computer Interface (BCI) system designed for hands-free computer and game control using affordable hardware. Developed as an undergraduate research project at the Department of Information Science and Engineering, Mysuru Royal Institute of Technology (MRIT), the system captures raw EEG signals from a single frontal dry electrode (Ups & Downs Labs BCI Kit) via an Arduino sampling at 512 Hz. Key components of the workflow include: Signal Processing: Implements a 50 Hz IIR notch filter and a 0.5–30 Hz fourth-order Butterworth band-pass filter to remove mains interference and baseline drift. Feature Extraction: Extracts eight spectral features per one-second window using Welch’s method, with the alpha/beta ratio serving as the primary indicator for cognitive states. Classification: Utilizes a Support Vector Machine (SVM) with an RBF kernel (trained on 4,668 windowed feature vectors) to classify user states as relaxed or attentive. Control & Interaction: Maps predicted mental states to keyboard inputs using pyautogui for real-time game control, complemented by a threshold-based jaw-clench detector for secondary input commands.