摘要
基于信息融合理论和线性鉴别分析,提出了一种改进的并行特征融合人脸表情识别方法。该方法首先将不同表征下的人脸表情特征利用复向量组合起来,构成复特征向量,然后利用具有不同权重的最大散度差鉴别分析方法进行进一步的复特征提取。在不同样本库、不同类型特征融合下的实验结果表明,该方法在优化投影轴和避免"小样本"问题的同时得到了满意的识别结果。
An improved maximum scatter difference discriminate criterion method based on information fusion theory is proposed in emotion recognition. Firstly, the complex feature vectors of different features are computed. Then, complex features are extracted by the maximum scatter difference discriminate criterion of different weight. Experiment results with different samples and features show the efficiency of the method and it can avoid the ＂Small Sample Size＂ and ＂inferior＂ problems. The correct recognition rate is further improved by the proposed feature fusion method.
出处
《广西大学学报（自然科学版）》
CAS
CSCD
北大核心
2009年第5期700-703,713,共5页
Journal of Guangxi University（Natural Science Edition）
基金
国家自然科学基金资助项目(60773113)
重庆市杰出青年科学基金(2008BA2041)
关键词
表情识别
特征融合
最大散度差鉴别分析
expression recognition
feature fusion
maximum scatter difference discriminate criterion
