A novel recognition method for electronic nose using SVM with compositional Gaussian kernel matrix

For an electronic nose (E-nose), training dataset is nonlinear due to the learning process, and the recognition accuracy of conventional classifier is not very ideal. An improved support vector machine (SVM) with compositional Gaussian kernel matrix is used as classifier of E-nose. The key point of this paper is to analyze the performance of SVM with the proposed kernel matrix and study how to obtain the best parameters of classifier. The results show that the recognition accuracy of SVM with 2-dimensional compositional Gaussian kernel matrix is 98.75% which is much higher than the other methods. Meanwhile, time consumption of this proposed method is much less than others. All results make it clear that SVM with compositional Gaussian kernel matrix is an ideal classifier for E-nose. © 2013 by Binary Information Press.