Hybrid deep models for parallel feature extraction and enhanced emotion state classification.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Machine Learning

Background:

• Emotions significantly impact human cognition and stress levels, making emotion classification crucial for AI and robotics.
• Traditional machine learning methods offer limited performance in emotion recognition from electroencephalogram (EEG) data.

Purpose of the Study:

• To develop and evaluate a hybrid deep learning model for precise emotion classification using EEG signals.
• To enhance the accuracy and efficiency of emotion recognition systems for advanced applications.

Main Methods:

• A hybrid deep learning architecture combining AlexNet and DenseNet for feature extraction.
• Feature fusion followed by Principal Component Analysis (PCA) for dimensionality reduction.
• Classification of reduced features using a multi-class Support Vector Machine (SVM).

Main Results:

• The model achieved high accuracy on the DEAP dataset (95.54% valence, 97.26% arousal) and EEG Brainwave dataset (98.42%).
• Demonstrated superior performance over existing methods in precision, recall, F1-score, specificity, and Mathew correlation coefficient.
• The hybrid approach effectively captures complex patterns in EEG data for reliable emotion detection.

Conclusions:

• The proposed hybrid deep learning model offers a significant advancement in EEG-based emotion classification.
• The model shows great potential for applications in human-computer interaction and mental health monitoring.
• This approach provides a robust and accurate method for understanding human emotional states.