Automatic Facial Expression Recognition Based on Hybrid Features

This paper describes an automatic recognition approach for facial expression from nearly front view face image. This work has two contributions. The first is a method for adaptively generating the initial model for Active Appearance Models (AAM) fitting, which allows the facial features under large variation to be detected more accuracy. The second is the introduction of a set of hybrid expression features, which consist of geometric features, AAM shape and appearance parameters. These features can be optimized when adopting quadratic mutual information (QMI) feature selection method. The experimental results on CAS-PEAL facial expression database show that by using support vector machine (SVM) classifier, the proposed method can achieve the recognition rate of 87.33%. (C) 2012 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Hainan University.