Toward emotion aware computing: an integrated approach using multichannel neurophysiological recordings and affective...

Area of Science:

• Neuroscience
• Affective Computing
• Biomedical Signal Processing

Background:

• Accurate classification of emotional states from neurophysiological data is crucial for understanding human responses.
• Current models often lack robustness or fail to account for the independent nature of emotional dimensions like arousal and valence.

Purpose of the Study:

• To propose a robust, gender-specific methodology for classifying four emotional states using electroencephalography (EEG) signals.
• To implement a two-step classification process based on the bidirectional emotion theory, separating arousal and valence discrimination.

Main Methods:

• Utilized EEG signals recorded during passive viewing of International Affective Picture System stimuli.
• Employed a two-step classification: first, arousal discrimination, then valence discrimination.
• Applied Mahalanobis distance (MD) and Support Vector Machines (SVM) classifiers.

Main Results:

• Achieved high classification rates: 79.5% for MD and 81.3% for SVM.
• Results significantly surpassed those of previous studies in emotion classification.
• Demonstrated the effectiveness of a gender-specific, two-dimensional approach to emotion classification.

Conclusions:

• The proposed methodology offers a robust and accurate approach to classifying emotional states from EEG data.
• This work represents a significant step towards objective emotional measurement for human-computer interaction applications.
• The gender-specific, bidirectional model advances the field of affective computing.