【摘要】 在电子鼻的模式识别方法中,主成分分析(PCA)是常用的方法之一。然而,主成分分析在计算过程中消除了各变量(对应于电子鼻的各测量传感器)问的相关性,这与传感器阵列的交叠感应特性相悖,致使时常无法正确鉴别多组分物品的类别。本文将Eilks准则引入主成分分析中,解决了酒类鉴别中主成分主轴向量的选择问题,实现了3种不同种类酒的正确鉴别,突破了纯粹的主成分分析模式。同时,指出了在用电子鼻鉴别多组分物品的种类时,主成分主轴的选择并非完全按照主成分贡献率的大小来确定的,这为今后的相关研究提供了一些有益的参考。更多还原

【Abstract】 Principal component analysis(PCA)is by far a popular method among pattern recognition algorithms of electronic nose.But the correlation among the variables that correspond to each gas sensor of electronic nose is elimi- nated in the calculation process of PCA.This violates the idea of making well use of cross-response of gas sensor ar- rays,so different kinds of products composed of many components can not be correctly discriminated and identified. In this paper,Wilks rule and PCA are combined together to solve the selection problem of principal vectors in drink identification,so that three kinds of drinks were accurately classified,which break the analysis mode of pure PCA. At the same time,it is pointed out that the principal vector selection is not determined totally according to the eigen- value corresponding to principal vector,which provides some useful guidance for similar investigation in the future.更多还原