Gender Difference in EEG Emotion Recognition with Overlapping Shifting Window

Gender in HCI has become crucial part due to the rising acknowledgment that computers must understand and adapt to the user's gender and emotional states. Hence, this work analyses the gender difference in emotion recognition based on the EEG signals. This paper used the overlapping shifting window mechanism to improve the emotion classification accuracy. Considering the frequency band in brain signals, we also investigate the critical frequency bands in alpha. Following that, we use PCA to reduce the dataset's dimensionality and utilize SVM to make a binary classification of valence and arousal emotions. We use a public dataset of EEG-based emotions comprising 13 female and 15 male subjects. According to the experiment results, the low $\alpha$ frequency (8-10 Hz) is more reliable for recognizing emotion. As for the epochs of shifting, the shorter the epochs window, the better the emotion classification accuracy. The average results of emotion classification accuracies in females reach 79.4% for valence and 78.05% for arousal, while the males obtain 81.7% for valence and 81.4% for arousal. Females are more affected by the valence emotion than by the arousal mood. In males, however, there is little difference between arousal and valence emotion perception. Furthermore, females have more complex aspects of valence and arousal emotion recognition than males.