一种基于Mahalanobis距离和主成分分析的电子鼻信号预处理方法

An Application of PCA-RBF Neural Network Based on Anomaly Detection for Electronic Nose Gas Detection.

针对样本集中的少数异常样本便可导致网络的稳定性下降甚至失效的问题。该文提出了基于马氏距离异常检测的PCA-RBF（Principal Component Analysis-Radial Basis Function）网络模型,将粗糙样本集经异常检验后进行PCA-RBF网络的识别。实验证明该方法能够克服异常样本的干扰,提高了网络的稳定性和识别能力。

Due to the fact that some anomaly of the sample set can decline the stability of the network.In this paper,an abnormal detection based on mahalanobis distance and a novel network PCA-RBF （Principal Component AnalysisRadial Basis Function） are proposed. PCA-RBF network identification is made after the anomaly detection. Experiment results show that this method can overcome the interference of abnormal samples and improve the network＇s stability and ability to identify.