AsEmo: Automatic Approach for EEG-Based Multiple Emotional State Identification.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Electroencephalogram (EEG) is crucial for emotion recognition via biometrics.
• EEG data analysis is challenging due to temporal variations and spatial-temporal correlations.
• Advanced methods are needed for accurate emotion state discrimination from EEG.

Purpose of the Study:

• To propose AsEmo, a new method for emotion recognition from multi-class EEG data.
• To extract effective features that improve classification performance for diverse emotional states.
• To develop a robust, subject-independent system for real-time emotion recognition.

Main Methods:

• AsEmo automatically determines optimal spatial filters using the explained variance ratio (EVR).
• It employs a subject-independent approach for real-time processing of emotion EEG data.
• The method is adaptable for both binary-class and multi-class classification tasks.

Main Results:

• AsEmo automatically identifies spatial filter coefficients for optimal feature extraction.
• The subject-independent technique ensures robustness for analyzing new data.
• Experimental results show AsEmo outperforms existing methods by 2-8% in classification accuracy.

Conclusions:

• AsEmo offers an effective and robust solution for EEG-based emotion recognition.
• The method's automatic feature extraction and subject-independent nature facilitate real-time applications.
• AsEmo demonstrates significant improvements in classification accuracy across various emotional states.