Analyzing Gaze Behavior During Emotion Recognition Using I-DT Fixation Detection

Analysis of gaze behavior during emotion perception using eye-tracking technology and fixation detection algorithms.

Overview

This study examines individual differences in gaze patterns across five observers viewing 15 standardized facial emotion images (3 identities × 5 emotions: Anger, Fear, Happy, Neutral, Sad). We apply the I-DT fixation detection algorithm to quantify eye movement metrics and spatial attention distribution.

Methodology

• Observers: 5 (anonymized as Observer_01 to Observer_05)
• Stimuli: 15 facial emotion images
• Recording Duration: 12 seconds per stimulus
• Display Resolution: 1920 × 1080 pixels
• Fixation Algorithm: I-DT with 60px dispersion & 100ms duration thresholds

Installation

git clone https://github.com/kidat/emotion-eyetracking-study.git
cd emotion-eyetracking-study
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

Usage

Data Collection

Run the eye-tracking experiment and record gaze data:

python emotion_eyetracking_study.py

This executes the experimental protocol:

• Stimulus presentation
• Eye-tracker calibration
• Real-time gaze recording
• Automatic CSV output to Output/ directory

Data Analysis

Run the comprehensive gaze analysis pipeline:

jupyter notebook dataAnalysis.ipynb

The notebook performs:

• Eye-tracking data loading and preprocessing
• I-DT fixation detection
• Fixation metrics computation
• Spatial attention heatmap generation
• Scanpath visualization
• Emotion-based statistical analysis
• Cross-observer comparative analysis

Analysis Output

• Fixation metrics (duration, count, saccade velocity)
• Spatial attention heatmaps
• Scanpath visualizations
• Emotion-based statistical summaries
• Cross-observer comparative analysis

Project Structure

├── emotion_eyetracking_study.py    # Data collection
├── dataAnalysis.ipynb              # Analysis & visualization
├── Data/                           # Raw stimuli & eye-tracking data
├── Output/                         # Results & processed data
└── README.md

Requirements

Python 3.7+, pandas, numpy, matplotlib, seaborn, scipy, opencv-python, jupyter

Technical Stack

• Eye Tracking: Hengam System
• Data Processing: Python, Pandas, NumPy
• Visualization: Matplotlib, Seaborn
• Analysis: SciPy, I-DT Algorithm

Ethical Statement

All observer data are anonymized and collected with informed consent according to research ethics guidelines.

Author

Kidu A. Welegerima

License

MIT License - See LICENSE file for details

---

Last Updated: December 2025