基于Gabor特征融合与LBP直方图的人脸表情特征提取方法  

Extraction method for facial expression features based on Gabor feature fusion and LBP histogram

作　　者：牛连强[1,2] 赵子天 张胜男[2] 

机构地区：[1]沈阳工业大学软件学院,沈阳110870 [2]沈阳工业大学信息科学与工程学院,沈阳110870

出　　处：《沈阳工业大学学报》2016年第1期63-68,共6页Journal of Shenyang University of Technology

基　　金：沈阳市科技计划项目(F12-168-9-00)

摘　　要：针对Gabor特征全局表征能力弱以及特征数据维数存在冗余的问题,提出了一种采用Gabor多方向特征融合与分块直方图相结合的方法以有效提取表情特征.通过对不同表情的重要特征部位进行细化,采用Gabor滤波器有针对性地提取相关区域的多尺度和多方向特征,并对同尺度的特征进行融合,利用各区域内融合特征的直方图分布来表征图像.该方法可以提高特征提取的准确性,有效突出重要特征的辨识作用,大幅度降低特征的维数,在JAFFE表情库可以达到100%的识别率.In order to solve the problem that the Gabor features exhiblt weak global expression ability and redundant feature data dimensionalities, a method in combination with both Gabor multi-direction feature fusion and block histogram was proposed to effectively extract the facial expression features. Through refining the important featute parts in different expressions, the multi-scale and multi-direction features in relevant areas were purposefully extracted with Gabor filters, and the features with the same scale were fused. In addition, the images were characterized with the histogram distribution for the fused features in various areas. The proposed method can improve the accuracy of feature extraction, effectively highlight the identification effect of important features, and greatly reduce the feature dimensionalities. Furthermore, the proposed method can reach 100% recognition rate in the JAFFE expression dataset.

关 键 词：表情识别 GABOR变换 特征融合 局部二进制模式 分块直方图 多尺度 多方向 维数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]