核广义多重集的典型相关分析  

Canonical Correlation Analysis of Kernel Generalized Multiset

作　　者：黄遵文 HUANG Zunwen(Ping An Kaicheng Intelligent Safety Equipment Co.,Ltd.,Anhui 232008,China.)

机构地区：[1]平安开诚智能安全装备有限责任公司,安徽232008

出　　处：《电子技术（上海）》2023年第6期328-330,共3页Electronic Technology

摘　　要：阐述典型相关分析(CCA)是一种重要的多特征提取方法,它是研究变量之间相关的一种统计分析方法,能够有效地揭示变量之间的相互线性依赖关系。传统CCA方法没有充分利用样本类别信息并且只能处理简单的线性问题,提出一种核广义多重集典型相关分析方法(KGMCCA),通过引入核方法和监督信息,能够提取更具鉴别性的特征信息。在人脸数据集上的实验结果表明,该方法相较于其他特征提取方法具有更好的识别率。This paper expounds that Canonical correlation analysis(CCA)is an important multi feature extraction method,which is a statistical analysis method to study the correlation between variables,and can effectively reveal the linear dependence between variables.The traditional CCA method does not make full use of the sample category information and can only deal with simple linear problems.It proposes a Kernel Generalized Multiset Canonical Correlation Analysis method(KGMCCA),which can extract more discriminative feature information by introducing the kernel method and supervisory information.The experimental results on facial datasets show that this method has better recognition rate compared to other feature extraction methods.

关 键 词：智能算法 核广义多重集 人脸识别 矩阵类内散布 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]