摘要
针对单幅表情图像识别缺乏表情间关联性及单分类器的局限性问题,提出一种基于反向协同显著区域特征的人脸表情识别方法.该方法首先对数据库进行预处理,获取表情图像的纯人脸区域,再选取相同的人七张不同表情图像,利用反向协同显著区域算法对选取图像提取表情之间的变化区域并作为显著区域,然后利用纹理和形状特征对显著区域进行描述,最后采用多分类器决策机制进行分类.在JAFFE和CK人脸表情库的实验结果表明,该方法在降低特征维度的同时,能挖掘出表情的显著区域部分并能对表情进行有效的描述,与其他近似的人脸表情识别方法对比,识别率平均提高了2.5%.
To solve the problems of absence of expression-correlation in single-image-based expression recognition,and limitation of one classifier,a novel facial expression recognition method based on reverse co-saliency region features is proposed in this paper. First,pure face regions of facial expression images are achieved to preprocess the database,then seven different classifications of expression images of a same person are used to extract facial changed regions of different expressions by a reverse co-saliency features method,named Reverse Co-Salient Regions（ RCSR）. Second,the extracted salient regions are described as texture and shape features LBP and HOG. Finally,based on RCSR features,multi-classifiers decision mechanism is used for expressions recognition. The proposed method is extensively evaluated on JAFFE database and CK database. The experimental results show that the proposed methods can reduce feature dimension and mining the salient regions and enhances the power of description facial expressions. In compared with other similarity methods in facial expression recognition,the average recognition rate of the proposed methods is increased by nearly 2. 5%.
作者
罗源
张灵
陈云华
曾碧
姜文超
LUO Yuan;ZHANG Ling;CHEN Yun-hua;ZENG Bi;JIANG Wen-chao(School of Computer, Guangdong University of Technology, Guangzhou 510006, China)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第7期1585-1589,共5页
Journal of Chinese Computer Systems
基金
广东省自然科学基金博士启动项目(2014A030310169)资助
广东省自然科学基金面上项目(2016A030313703)资助
广东省自然科学基金(2016A030313713)资助
广东省科技计划项目(2016B030305002)资助
广东省交通运输厅科技项目(科技-2016-02-030)资助
广州市重点科技项目(201604020016)资助
关键词
反向协同
显著区域特征
多分类决策机制
表情识别
reverse co-saliency
salient region features
expression-correlation multiple classifiers decision
facial expression recognition
