基于粗糙集和集成剪枝的人脸表情识别方法  

A facial expression recognition method based on rough set and integrated pruning

作　　者：唐玉梅 李丹杨 吴亚婷 黄仕松 陈星 吴义青 TANG Yumei;LI Danyang;WU Yating;HUANG Shisong;CHEN Xing;WU Yiqing(College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州大学大数据与信息工程学院,贵阳550025

出　　处：《智能计算机与应用》2023年第4期20-26,共7页Intelligent Computer and Applications

摘　　要：人脸表情识别在疲劳驾驶监测等场景有着广泛的运用。针对人脸表情识别难度大,单一分类器泛化能力较弱的缺点,基于集成学习理论,提出一种基于粗糙集和集成剪枝的人脸表情识别方法。首先,更改卷积神经网络参数得到若干基分类器;其次,结合粗糙集理论,根据基分类器的预测结果构建信息决策表,将分类器选择转化为知识约简过程,剔除系统中弱分类器或冗余分类器,选出分类器子集;最后,用大多数投票法将选择出来的分类器子集组合。和多个集成剪枝算法对比,本文集成剪枝算法在表情数据集上具有较高的识别准确率。Facial expression recognition is widely used in fatigue driving monitoring and other scenes.Aiming at the disadvantages of difficult face expression recognition and weak generalization ability of single classifier,a face expression recognition method based on rough set and integrated pruning is proposed based on ensemble learning theory.Firstly,the convolutional neural network parameters are changed to obtain some base classifiers.Secondly,combined with rough set theory,an information decision table is constructed according to the prediction results of base classifier,and classifier selection is transformed into a knowledge reduction process.Weak or redundant classifiers are removed from the system,and classifier subsets are selected.Finally,the subsets of selected classifiers are combined using majority voting methods.Compared with multiple integrated pruning algorithms,the integrated pruning algorithm in this paper has higher recognition accuracy in facial expression data sets.

关 键 词：表情识别 卷积神经网络 集成剪枝 粗糙集理论 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]