| Abstract |
| Electroencephalography (EEG)-based emotion recognition has attracted increasing attention because of its ability to capture intrinsic affective states beyond any observable behavior. However, the effective representation of high-dimensional EEG features and the lack of age-specific analyses remain major challenges in fine-grained emotion recognition. To address these limitations, we analyzed how EEG structural features vary across different age groups based on the EEGEmotions-27 dataset. A substantial overlap was seen among the top-ranked EEG features between the young and old age groups. Furthermore, suppressed feature analysis suggested that performance degradation was more strongly associated with a reduction in dimensionality than with a clear mismatch between age-specific feature set. These findings provide preliminary evidence that age-related variations may be more common with the relative importance of shared EEG features rather than in the presence or absence of entirely distinct feature subsets. This analysis may contribute to developing interpretable, lightweight, and age-specific EEG-based emotion recognition systems. |
|
|
| Key Words |
| EEG, Emotion Recognition, Fine-grained Emotion Classification, Age-Dependent Analysis |
|
|
 |
|
