CTFear: A Fear Emotion Intensity Classification Method Based on EEG and Real-Time Labeling in Virtual Environment

Area of Science:

• Neuroscience
• Virtual Reality Technology
• Psychological Measurement

Background:

• Fear significantly impacts human behavior and psychological states, but quantifying its intensity is difficult.
• Virtual reality (VR) offers a controlled environment for psychological interventions like exposure therapy.
• Accurate real-time fear intensity measurement is crucial for understanding emotional responses and developing effective therapies.

Purpose of the Study:

• To propose a novel labeling paradigm for real-time, continuous subjective fear intensity measurement in immersive virtual environments.
• To develop and validate a deep learning model (CTFear) for decoding fear intensity from electroencephalography (EEG) signals.
• To enhance fear intensity labeling accuracy using haptic feedback as physical anchors.

Main Methods:

• Users labeled their subjective fear intensity in real-time by controlling a VR controller trigger while watching immersive VR videos.
• Haptic vibration cues were integrated at specific trigger depths to serve as physical anchors for fear level discrimination.
• A deep learning model, CTFear, combining convolutional neural networks and Transformers with topology-aware spatial positional encoding, was designed for EEG signal analysis.

Main Results:

• CTFear achieved high average F1 scores in classifying fear intensity: 0.86 (two-class), 0.76 (three-class), and 0.67 (four-class) in cross-trial validation.
• Cross-subject validation yielded F1 scores of 0.80 (two-class), 0.64 (three-class), and 0.55 (four-class).
• The proposed method significantly outperformed existing approaches in multi-class fear intensity classification tasks.

Conclusions:

• The CTFear model effectively decodes fear intensity from EEG signals, demonstrating its potential for objective emotion assessment.
• The VR-based labeling paradigm with haptic feedback provides a reliable method for capturing subjective fear intensity.
• This research offers a promising pathway for real-time emotion monitoring and human-computer interaction in virtual environments.