基于KPCA和RBF网络的电子鼻气体识别

An electronic nose designed for gas recognition based on the kernel principal component analysis and RBF neural network

提出了将KPCA特征提取和RBF网络识别相结合的气体检测方法,设计了一种用于气体实时检测的电子鼻系统,探讨了核主成分分析(KPCA)和RBF神经网络相结合进行气体识别的可行性。将传感器阵列的动态检测方法应用到电子鼻系统中,对甲苯、乙酸酐、乙醚、丙酮4种气体进行检测,针对响应信息的非线性变化利用KPCA进行特征提取,并作为RBF网络的输入,检测系统重复性和稳定性好,识别率可达87.5%。

This paper presents a gas detection method based on KPCA and RBF （radial basis function） network. An on-line electronic method and recognition can be achieved correctly. Dynamic detection is used to test toluene, acetic anhydride, aether and acetone. In view of the non-linear variation in the response information, the KPCA is used to extract the feature as the input of the RBF network. The results indicate that the electronic nose system good in repeatability and reliability. The recognition rate comes to 87.5%.