期刊名称：Journal of Computational Science and Technology
电子版ISSN：1881-6894
出版年度：2008
卷号：2
期号：4
页码：401-412
DOI：10.1299/jcst.2.401
出版社：The Japan Society of Mechanical Engineers
摘要：Automatic recognition of facial expressions can be an important component of natural human-machine interactions. While a lot of samples are desirable for estimating more accurately the feelings of a person (e.g. likeness) about a machine interface, in real world situation, only a small number of samples must be obtained because the high cost in collecting emotions from observed person. This paper proposes a system that solves this problem conforming to individual differences. A new method is developed for facial expression classification based on the combination of Holographic Neural Networks (HNN) and Type-2 Fuzzy Logic. For the recognition of emotions induced by facial expressions, compared with former HNN and Support Vector Machines (SVM) classifiers, proposed method achieved the best generalization performance using less learning time than SVM classifiers.
