Study of Recognizing Spontaneous Facial Expressions of a Person who Watches Web News Based on ASM and Bayesian Network

摘要

To achieve a comment system that can automatically estimate emotions of a person who watches web news, this paper proposes a computer vision based method that can recognize facial expressions generated in response to Web news. During the learning session, first, ASM (Active Shape Model) is applied to the face so that some feature points are located. From the located feature points, some features such as distances between feature points are obtained. The features collected from sample data are used for nodes of Bayesian Networks. During facial expression estimation session, the features that are computed from the input facial video sequence by the same procedure as the learning session are input to the Bayesian Networks so that the recognition result is obtained as the facial expression category that maximizes the a-posteriori probability. Experimental results demonstrate the effectiveness of the proposed method.