基于逐步判别分析和BP神经网络的电子.PDF

传 感 技 术 学 报 第 23 卷 第 10 期 Vol． 23 No． 10 2010 年 10 月 CHINESE JOURNAL OF SENSORS AND ACTUATORS Oct． 2010 23 Discrimination of Different Storage Time of Pork by Electronic Nose* ， * HONG Xuezhen WANG Jun ? ， ， ， ? Department of Biosystems Engineering Zhejiang University Hangzhou 310029 China ? ， Abstract In this study a rapid detection method based on electronic nose for pork freshness was developed． An ? ， ? ? ? ， electronic nose E-nose PEN 2 was employed to classify pork groups with different storage time 0 ～ 7 d 42 sam- ， ples were tested everyday． The mass of each sample was 10 g and the headspace-generated time was 5 min． The 60 th data from the response of the E-nose was extracted for further analysis． After employing the Linear Discriminant ? ? ， Analysis LDA samples could be well classified according to their storage time． Stepwise Linear Discriminant A- ? ? ? ? nalysis Step-LDA and Back Propagation Neural Network BPNN were also employed to predict the storage time of ， the samples． The result showed that Step-LDA got 100% training accuracy with 97． 92% prediction accuracy and BPNN got 94． 17% training accuracy with 93． 75% prediction accuracy． This study implied that electronic nose