Multi-Class Emotion Recognition Within the Valence-Arousal-Dominance Space Using EEG.

Emotions play an important role in our daily life. Detecting emotions is a natural aspect of human communication and seems like a matter of course to us. However, the brain's procedure of emotion processing is far more complex. We know little about the mechanisms of emotion regulation. Due to that, emotion recognition based on EEG brain signals has drawn much attention in research. This paper presents an approach for modeling and classifying different emotions within the Valence-Arousal-Dominance model. We investigate the effectiveness of high gamma frequencies (50-100 Hz) for emotion detection by dividing the standard EEG bandwidths into fine-graded point spectra, each with a span of 0.5 Hz. With an F1-score of 99.79%, we not only show that our method is very well suited for discriminating different emotional states, but we also identify the most important high gamma frequency sub-bands. Our findings are consistent with other studies and further suggest the high gamma activity of the brain in emotion processing.