H-036 Subtle Facial Expression Recognition Based on Expression Category-dependent Motion Magnification

概要

論文の詳細を見る
Previous subtle facial expression recognition studies have primarily focused on extracting the subtle changes of facial expression but not classifying the extracted subtle expression into categories such as anger and happiness. In this paper, we propose a novel method based on expression category-dependent motion magnification to recognize subtle facial expressions. Because it's difficult to discriminate subtle expressions directly from raw subtle facial deformation data, we use expression category-dependent motion magnification to reliably transform subtle facial expressions into the corresponding exaggerated facial expressions. The magnified facial expressions become much more discriminable and we can infer the expression category of the subtle expression from them. Experiments using our database showed that the proposed method has a promising performance on subtle expression recognition, especially in the early stage of making facial expression.
FIT(電子情報通信学会・情報処理学会)推進委員会の論文
2010-08-20