Automated Emotion Identification Using Fourier-Bessel Domain-Based Entropies

Area of Science:

• Physiological computing
• Affective computing
• Machine learning for emotion recognition

Background:

• Increasing human-computer interaction necessitates dynamic and contextual systems.
• Emotion recognition requires understanding user's emotional state via physiological signals.
• Electrocardiogram (ECG) and electroencephalogram (EEG) are key physiological signals for emotion detection.

Purpose of the Study:

• To propose novel entropy-based features in the Fourier-Bessel domain for emotion recognition.
• To utilize Fourier-Bessel series expansion (FBSE) for representing non-stationary physiological signals.
• To develop and evaluate a machine learning model for emotion detection using these features.

Main Methods:

• Decomposition of ECG and EEG signals into narrow-band modes using FBSE-based empirical wavelet transform (FBSE-EWT).
• Computation of proposed entropy features for each mode to create a feature vector.
• Development of machine learning models, specifically K-nearest neighbors (KNN), for classification.

Main Results:

• The proposed algorithm achieved high accuracies: 97.84% for arousal, 97.91% for valence, and 97.86% for dominance.
• Novel entropy features in the Fourier-Bessel domain demonstrated superior performance compared to traditional methods.
• FBSE proved effective in representing non-stationary physiological signals like ECG and EEG.

Conclusions:

• The developed entropy features are highly suitable for accurate emotion recognition from physiological signals.
• The FBSE-EWT approach provides a robust method for feature extraction from non-stationary biosignals.
• This research contributes to more dynamic and context-aware human-computer interaction systems.