(英) |
The classification of facial expressions by linguistic categories, which is currently widely used in facial expression recognition, has difficulty in describing ambiguous facial expressions. In response to this, the dimensionality theory for continuously changing facial expressions has been attracting attention, and many studies have been conducted on the psychological space of facial expressions obtained by the SD method. However, the psychological space has the problem that it is difficult to link physical stimuli and psychological characteristics. Recently, Sumiya et al. proposed a psychophysical space of facial expressions that introduces a discrimination threshold as a psychological property to the image space of facial expressions as physical stimuli. On the other hand, the dimensionality of the expression image space is so high that it is important to determine the dimensionality of the expression manifold. In this study, we estimated the high-dimensional expression discrimination threshold ellipse and studied to determine the effective dimension. |