摘要
储粮的霉变与否大部分靠人的嗅觉感受评定,判定结果的准确性难以保证。针对这一问题,利用自行研制的一组厚膜金属氧化锡气体传感器(MOS)阵列,获得了传感器与粮食挥发气味的整个反应过程的数据,提取了传感器与气味反应的特征值。用常规的主成分分析对提取的特征值进行分析,结果表明很难区分,而用RBF神经网络进行分析,回判率达到100%,识别正确率达到92 19%。
The foodstuff is moldy or not often evaluated by people's nose,the accuracy of evaluated results is hard to be guaranteed.To the problem,used a thick tin oxide gas sensor array that was designed by themselves,picked up the feature parameters of the sensor acted with gas after getting the whole acting process data.Principal component analysis(PCA)and artificial neural network(ANN)were used to analysis the feature parameters,PCA was failure and ANN was successful.The recognition probability of the ANN is 9219%.
出处
《仪表技术与传感器》
CSCD
北大核心
2005年第3期51-52,共2页
Instrument Technique and Sensor
基金
国家社会公益研究专项资金(2001DTA40038)
关键词
气敏传感器阵列
特征提取
主成分分析
神经网络
Gas Sensor Array
Character Pick-up
Principal Component Analysis
Neural Network
