Emotion Recognition in Gaming using EEG Signals

Overview

This project applies machine learning to EEG (Electroencephalography) data from 27 participants to classify emotional states.
The task was simplified into binary classification — grouping Calm, Satisfaction, and Funny as positive emotions, and Boredom and Horrible as negative emotions.

We tested SVM, Logistic Regression, and Random Forest, with Random Forest achieving the best performance.

---

Dataset

• Source: Cleaned EEG CSV files from 27 participants, 4 games each (5 minutes per game)
• Features: Averaged EEG channels by brain region (frontal, temporal, parietal, occipital) plus mean and variance
• Labels: Binary emotion labels — Positive (Calm, Funny, Satisfaction) and Negative (Boredom, Horrible)

---

Preprocessing

1. Combined CSVs into a single DataFrame
2. Standardized column names and units
3. Added averaged brain region features
4. Converted multi-class emotions into binary labels
5. Split into 80% training and 20% testing

---

Models Tested

• SVM — Moderate performance, struggled with high-dimensional space
• Logistic Regression — Simple and interpretable, but underperformed
• Random Forest — Best performance, handled non-linear relationships well
  • Feature importance: Spread across multiple brain regions rather than concentrated in a few channels

---

Results

• Accuracy: 62.9%
• Confusion Matrix: [[275405 177988] [113154 218837]]