Classification of Chinese drinks by a gas sensors array and combination of the PCA with Wilks distribution

The classification for three kinds of Chinese drinks is made based on a more in-depth investigation in this paper using with a gas sensor array composed of six Taguchi gas sensors (TGS) and principal component analysis (PCA). Because the first two principal components employed as analysis vectors cannot carry out the drinks classification task, we propose a new selection method of principal components by means of Wilks Λ-statistic. Using the method, the fourth and fifth principal components are selected as analysis vectors and the drink classification task is well achieved. At the same time, a feature extraction method of responses of the gas sensor array is also presented. The feature extraction method is very straightforward and the feature parameters possess clear physical meanings and represent mainstream traits of sensor response. So we think a new idea may emerge especially in respect of how to utilize well the PCA in the future works and the feature extraction method may be widely applied in the field of gas sensors.