Fractal Analysis of Electrodermal Activity for Emotion Recognition: A Novel Approach Using Detrended Fluctuation...

Area of Science:

• Affective Computing
• Physiological Signal Processing
• Biomedical Engineering

Background:

• Emotion recognition from physiological signals is crucial for mental health and human-computer interaction.
• Electrodermal activity (EDA) is a key physiological signal for emotion detection.
• Existing methods often focus on simple EDA features, potentially missing complex signal dynamics.

Purpose of the Study:

• To introduce and evaluate a novel approach for emotion recognition using fractal analysis of EDA signals.
• To explore the utility of fractal features in capturing multi-scale dynamics of EDA.
• To assess the performance of fractal features in classifying emotional states and dimensions.

Main Methods:

• Fractal analysis techniques including detrended fluctuation analysis (DFA), Hurst exponent estimation, and wavelet entropy.
• Application of these methods to EDA signals from the CASE dataset.
• Machine learning models for emotion classification using extracted fractal features.

Main Results:

• Significant differences in fractal features were observed across five distinct emotional states.
• Wavelet entropy-derived features showed particular promise in distinguishing emotions.
• Fractal features demonstrated robust correlations with both arousal and valence dimensions of emotion.
• Machine learning classification achieved 84.3% accuracy and an F1 score of 0.802 using fractal features.

Conclusions:

• Fractal analysis effectively captures the complex, multi-scale dynamics of EDA signals for emotion recognition.
• This approach enhances the understanding of EDA's role beyond just arousal indication.
• The findings open new possibilities for developing advanced emotion-aware systems and affective computing applications.