Automatic EEG-based dream-related emotion recognition using fuzzy entropy and efficient signal decomposition methods.

Area of Science:

• Neuroscience
• Computational Psychology
• Signal Processing

Background:

• Dreams are linked to emotional processes and profound needs.
• Electroencephalogram (EEG) signals during Rapid Eye Movement (REM) sleep can be used to study dream emotions.
• The Dream Emotion Evaluation Dataset (DEED) provides a resource for this research.

Purpose of the Study:

• To develop an automated framework for classifying dream emotions (positive, neutral, negative).
• To utilize EEG signals captured during REM sleep for dream emotion analysis.
• To evaluate the performance of different signal decomposition and feature extraction techniques.

Main Methods:

• EEG signal decomposition using Discrete Wavelet Transform (DWT) and Empirical Mode Decomposition (EMD).
• Extraction of Fuzzy Entropy (FuzzEn) as a nonlinear feature from subbands.
• Feature selection using the ReliefF algorithm.
• Classification using K-Nearest Neighbors (KNN), Support Vector Machine, Extreme Gradient Boosting, and Random Forest.

Main Results:

• The EMD-FuzzEn method combined with KNN achieved superior classification performance.
• Accuracy rates included 92.33% for multi-class, 96.47% for neutral vs. non-neutral, and 90.69% for positive vs. negative emotions.
• The ReliefF algorithm identified temporal and frontal EEG regions (T7, T8 channels) as highly discriminative.

Conclusions:

• The proposed methodology shows strong potential for identifying dream emotions.
• The study highlights the efficacy of signal decomposition and feature extraction techniques.
• This work represents a substantial advancement in automated dream emotion classification.