基于LBP特征和稀疏表示的新生儿疼痛表情识别  

Recognition method for neonatal pain expression based on LBP feature and sparse representation

作　　者：卢官明[1] 石婉婉 李旭[1] 李晓南[2] 陈梦莹[2] 刘莉 

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003 [2]南京医科大学附属南京儿童医院,江苏南京210008 [3]瑞典于默奥大学应用物理与电子系,swedenumes90187

出　　处：《南京邮电大学学报（自然科学版）》2015年第1期19-25,共7页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(61071167;81202222);江苏省高校优秀中青年教师和校长境外研修计划;Swedish International Collaborative Research Grant(2011-7270)资助项目

摘　　要：面部表情被认为是新生儿疼痛评估的可靠指标。文中提出一种基于加权局部二元模式(LBP)特征描述符和稀疏表示分类器的新生儿疼痛表情识别方法。首先,经归一化后的面部图像采用一个特征向量来描述,这个特征向量是通过串接组合所有局部图像块的LBP特征图的加权直方图所得到的直方图序列。然后,采用主成分分析(PCA)方法对训练样本及测试样本图像的特征向量进行降维。最后,采用基于稀疏表示的分类器将测试样本图像的表情分为4类:平静、哭、轻度疼痛、剧烈疼痛。文中研究目的是通过利用基于计算机的图像分析技术来辅助临床医生评估新生儿疼痛。在新生儿面部图像数据库上进行的实验结果表明了该算法的有效性,表情分类的平均识别率高达84.50%。Facial expressions are considered as a reliable indicator in neonatal pain assessment. This paper proposes a novel recognition method for neonatal pain expression. The method can utilize the feature descriptors based on the weighted local binary pattern （LBP） and the classifier based on sparse representation. Firstly, the normalized facial image is described using a feature vector, which is histogram sequence obtained by concatenating the weighted histograms of the LBP feature maps of all the local blocks. Then, the principalc component analysis （PCA） method is used to reduce the dimensions of the feature vector of training and test samples. Finally, the over-complete dictionary is built and the classifier based on sparse representation is used to classify test samples into four classes of facial expressions： calm, crying, mild pain, and severe pain. The objective of this study is to assist the clinicians in assessing neonatal pain by utilizing computer-based image analysis techniques. Experimental results on neonate facial image database

关 键 词：表情识别 新生儿 疼痛表情 局部二元模式 稀疏表示 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]