摘要
针对目前多数表情识别算法都是基于浅层特征的,很难达到良好的识别效果,并且核主成分分析网络(PCANet)网络存在提取到的表情特征维数比较高致使识别时间较长和分类效率较低的问题,受到深度学习模型PCANet的启发,提出了一种结合核主成分分析网络(KPCANet)和线性判别分析(LDA)的表情识别算法.首先,利用基于KPCANet模型获取训练样本及测试样本的深层特征;然后,用LDA监督层对KPCANet模型获取的深层特征对表情图像特征进行监督投影,从而使表情特征具有类别区分性;最后,将经LDA投影的特征矩阵输入支持向量机(SVM)中对表情特征进行训练和分类.提出的KPCANet-LDA算法模型在人脸表情数据库CK+和JAFFE上进行实验,实验结果表明提出的算法具有良好的鲁棒性且识别率高于其他对比算法.
The most existing methods of facial expression recognition only extract low-level features,can not obtain satisfactory recognition results.The high dimension of features extracted by the principal component analysis network(PCANet)leads to high memory consumption and low classification efficiency.For solving these problems,inspired by PCANet model,a facial expression recognition method combining KPCANet(kernel principal component analysis network)and LDA(linear discriminate analysis)was proposed.Firstly,KPCANet model was used to get the high-level features of training samples and testing samples.Then,the LDA matrix was used to project the KPCANet features into a low-dimensional space to make the projected features more distinguishable for classification.Finally,the feature matrixs projected by LDA were used to train support vector machine(SVM)and classify the expression images.This proposed algorithm was tested on CK+and JAFFE databases of facial expression.The experimental results show that the proposed algorithm has good robustness and higher recognition rate.
作者
陈香敏
柯丽
杜强
CHEN Xiangmin;KE Li;DU Qiang(College of Electrical Engineering,Shenyang University of Technology,Shenyang 110870,China;Department of Information and Control Engineering,Shenyang Institute of Science and Technology,Shenyang 110167,China)
出处
《华中科技大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2020年第9期95-99,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(51377109)
辽宁省自然科学基金资助项目(2019-ZD-0204)。
关键词
表情识别
核主成分分析网络
特征提取
线性判别分析
深度学习
facial expression recognition
kernel principal component analysis network
feature extraction
linear discriminate analysis
deep learning