基于主成分分析和BP 神经网络的气体识别方法研究

本文将主成分分析法与BP算法相结合应用于气体传感器阵列信号的处理，并以一个由4个SnO2气体传感器组成的阵列为例，对其受到不同浓度的汽车、酒精二元气体的响应信号进行了分析，结果表明，主成分分析能够在保留测试数据量大量信息的前提下，给数据有效降维和预分类，以消除样本间的相关性。然后，再将所产生的新的样本空间作为BP网络输入，使之减少网络的输入数，简化网络结构，并在保持相同正确率的前提下，大大提高网络的学习速率。

Study on Gas Recognition Method with Principal Component Analysis and Back-Propagation Neural Network

Principal Component Analysis (PCA) and Back Propagation (BP) neural network were used for processing signals of a gas sensor array consisted of 4 SnO\-2 gas sensors. The responses of the array to gasoline, ethanol, and mixture of gasoline and ethanol were analyzed with the PCA and BP in series. The results showed that PCA can give a good classification for the gases, reduce the dimension of data, and eliminate the data correlation with retaining the most information. The new sample space produced by PCA is employed as the inputs of BP network. Hence, to the same recognition right rate, the construction of BP network is simplified, and the convergent speed of the BP network is enhanced greatly due to the number of the BP inputs is reduced.