Emotion Recognition from Non-Frontal Facial Images
Emotion Recognition from Non-Frontal Facial Images
复制标题
DOI：
10.1002/9781118910566.ch8
复制
发表时间：
2015-01
期刊：
影响因子：
--
通讯作者：
Wenming Zheng;Hao Tang;Thomas S. Huang
中科院分区：
文献类型：
--
作者：
Wenming Zheng;Hao Tang;Thomas S. Huang
本章综述了当前非正面面部情绪识别方法的最新进展，包括三维面部表情识别和多视角面部表情识别。首先简要介绍了人脸情感识别的方法。然后简要回顾了非正面面部情感识别研究中常用的面部表情数据库。利用三维人脸模型进行人脸识别的方法，由于其对姿态、尺度和光照变化的鲁棒性，已被证明比二维人脸图像具有更好的识别效果。为了进行情绪分类，可以选择一个分类器，例如支持向量机或Adaboost，然后根据几何特征将每个3D人脸模型分类到一个基本的情绪类别中。
This chapter reviews the recent advances of the current non‐frontal facial emotion recognition methods, including the three‐dimensional (3D) facial expression recognition and multiview facial expression recognition. It first gives a brief introduction of the facial emotion recognition methods. The chapter then briefly reviews the facial expression databases that are commonly used for the non‐frontal facial emotion recognition researches. The method of using of 3D face model for face recognition has been proven to achieve better recognition results than 2D facial images due to its robustness to the poses, scales, and lighting variations. To perform the emotion classification, one may choose a classifier, for example, the support vector machine or Adaboost, and then classify each 3D face model into one of the basic emotion categories based on the geometric features.