摘要
利用13个TGS系列的金属氧化物气敏传感器组成的阵列对3种乳制品进行了测量。提取每个传感器响应值的方差为特征值,借鉴逐步判别分析的思想,依据Wilks统计量最小的原则,逐步引入对分类有用的变量,实现了对传感器阵列的优化。对阵列优化前后的数据,用主成分分析(PCA)和Fisher判别分析(FDA)的方法进行对比研究。结果表明:使用优化后的传感器阵列,不仅主成分分析能将不同种类的乳制品很好的区分开;而且用Fisher方法时,交叉确认正确率也得到提高。因此,所给出的阵列优化方法是有效的。
Three kinds of dairy products were measured by a gas sensor array composed of thirteen sensors. Variances of thirteen sensor＇s signals were picked up as feature parameters. Using the idea of step-wise discriminant analysis, the useful variables were step by step contained in the array by the minimal Wilks A statistic, so the sensor array could be optimized. Principal Component Analysis （PCA） and Fisher discrim- inant analysis （FDA） were employed to treat with the test data of primary and optimized sensor array. For the optimized sensor array , the result of PCA indicated that three kinds of dairy products were discrimina- ted well; at the same time, the ratio of cross-validation of FDA was remarkably improved. Therefore the optimization method for sensor array given by this paper is effective.
出处
《传感技术学报》
CAS
CSCD
北大核心
2008年第7期1124-1127,共4页
Chinese Journal of Sensors and Actuators
基金
河南省杰出青年科学基金资助项目(0612000400)
关键词
电子鼻
传感器阵列
主成分分析
FISHER判别分析
electronic nose
sensor array
principal component analysis
fisher discriminant analysis
