Abstract
Facial impressions sometimes change when one's personality is understood in more depth. This study is aiming to analyze how to recognize facial expressions as shapes, and to suggest a computational addition method of facial impressions. First, we used PCA to feature points of faces, and reduce shape data sets to lower dimensions. Secondly, we asked subjects to recreate facial impressions by controlling principal component scores, and analysis a relationship between the scores and the impressions. Finally, we set facial impression axes, and propose a method for adding impressions to an input face shape.
