主成分分析在电子鼻鉴别分析中存在的缺陷及其改进方法

The Defects and Improving Methods of Principal Component Analysis in Identification Analysis based on Electronic Nose

主成分分析(PCA,Principle Component Analysis)在电子鼻鉴别分析中是一种常用的线形判别方法,然而,当所测样品等级质量差别很小,即样品挥发物成分基本接近时,电子鼻中各传感器所能反映样品差异的响应信息存在较大的重叠性或相关性,用传统的累计贡献率来选取前两个主成分进行鉴别分析效果往往不佳。本文从PCA降维的数学原理和传感器阵列特点出发,分析当样品等级质量差别很小、电子鼻各传感器响应信号重叠较大时这种选取主分量方法所存在的问题,在此基础上结合Wilks准则提出了选取主分量的新方法。实例证明了所提出的主成分选取方法是有效的。

Principal Component Analysis （PCA） is a popular method for linear discriminant analysis in the identification analysis of electronic noses. However, when the measured samples have smaller differences in quality, namely, the volatile components of samples are almost same, the response information of the electronic sensors has greater superposition and relativity. The identification effects using the first two principal components by traditional cumulative contribution are often poor. In this paper, the main defects of PCA method are discussed based on the mathematical principle of PCA and the characteristics of sensor array under this case. At the same time, a new method of selecting principal components combined with Wilks rule is given and proved effectively by example.