摘要
为了评估恶臭对人类和环境的影响,需要有效鉴别出恶臭成分,设计并研发了以传感器阵列为核心的在线检测恶臭成分的电子鼻系统。针对传感器响应曲线几何特征提取方式的不足,提出分段拟合曲线的方式进行特征值提取。采用主成分分析(PCA)和线性判别分析(LDA)两种降维方式对原始数据降维,并结合支持向量机(SVM)和BP神经网络进行分类识别。结果表明,SVM和BP神经网络结合LDA都能100%识别出三种恶臭物质,SVM结合PCA识别率为92%,BP神经网络结合PCA识别率为94%。
In order to evaluate the impact of foul smell on human-beings and environment,it is necessary to effectively identify foul smell’s components.An electronic nose system with sensor array as its core for the online detection of foul smell’s components is designed and developed.As for the insufficiency of the geometric feature extraction method of sensor response curve,the method of segment fitting curve is proposed to extract feature values.The principal component analysis(PCA)and linear discriminant analysis(LDA)are used for the dimension reduction of original data,and the support vector machine(SVM)and BP neural network are combined with them for the classification and identification.The results show that both SVM and BP neural network combined with LDA can recognize three kinds of foul smell substances in 100%,the recognition rate of SVM combined with PCA is 92%,and the recognition rate of BP neural network combined with PCA is 94%.
作者
刘伟玲
吴龙焦
张思祥
闫子琪
LIU Weiling;WU Longjiao;ZHANG Sixiang;YAN Ziqi(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300130,China)
出处
《现代电子技术》
北大核心
2020年第6期57-60,共4页
Modern Electronics Technique
基金
国家重大科学仪器设备开发专项项目(2012YQ060165)。
关键词
电子鼻系统
气体识别
特征值提取
特征降维
气体分类
在线检测
electronic nose system
gas identification
feature value extraction
feature dimension reduction
oder classification
only detection
