基于电子舌的天然药物的四类味觉分类辨识研究  

Study on four kinds of taste classification and identification of natural medicines based on electronic tongue

作　　者：王艳丽[1,2] 陈鹏举[2] 桂新景 姚静[1,3,4,5] 张璐 施钧瀚[1,3,4,5] 张耀 王柯涵[1,2] 李学林 刘瑞新 WANG Yan-li;CHEN Peng-ju;GUI Xin-jing;YAO Jing;ZHANG Lu;SHI Jun-han;ZHANG Yao;WANG Ke-han;LI Xue-lin;LIU Rui-xin(The First Affiliated Hospital of Henan University of Chinese Medicine,Zhengzhou 450000,China;Henan University of Chinese Medicine,Zhengzhou 450046 China;Level Three Laboratory of Chinese Traditional Medical Preparation of National Administration of Traditional Chinese Medicine,Zhengzhou 450000,China;Henan Modern Engineering Research Center for Clinical Application of Chinese Herbal Pieces,Zhengzhou 450000 China;Co-construction Collaborative Innovation Center For Chinese Medicine and Respiratory Diseases by Henan&Education Ministry of China,Henan University of Chinese Medicine,Zhengzhou 450000,China)

机构地区：[1]河南中医药大学第一附属医院,郑州450000 [2]河南中医药大学,郑州450046 [3]国家中医药管理局中药制剂三级实验室,郑州450000 [4]河南省中药饮片临床应用现代化工程研究中心,郑州450000 [5]河南中医药大学呼吸疾病中医药防治省部共建协同创新中心,郑州450000

出　　处：《中华中医药杂志》2021年第1期423-433,共11页China Journal of Traditional Chinese Medicine and Pharmacy

基　　金：国家自然科学基金面上项目(No.81774452,No.81773892);国家自然科学基金青年科学基金项目(No.81001646);河南省中医管理局中医药科学研究专项(No.2017ZY2039,No.2018ZY2131);河南省中医药拔尖人才培养项目专项课题(No.2019ZYBJ07);河南省高层次人才特殊支持“中原千人计划”——“中原青年拔尖人才”项目(No.ZYQR201912158);河南省卫生健康中青年学科带头人专项。

摘　　要：目的:为探究基于电子舌进行具有不同味道的天然药物区分的可行性及不同类型电子舌区分效果差异。方法:本实验采用TS-5000Z和α-ASTREE两种类型电子舌对分别具有酸、苦、甜、咸4种味道的共35种代表性天然药物及12种食品类成分进行检测,通过提取的电子舌传感器数据以及经典人群口尝评价(THTPM)结果,分别利用判别分析(DA)、最小二乘支持向量机(LS-SVM)、主成分分析-判别分析(PCA-DA)、偏最小二乘-判别分析(PLS-DA)4种算法建立四味的定性判别模型。结果:其中基于TS-5000Z型电子舌数据的上述4种算法的四味辨识准确率(交互验证)分别为72.34%、72.34%、76.60%、61.70%;基于α-ASTREE型电子舌数据的4种算法的四味辨识准确率(交互验证)分别为91.49%、89.36%、93.62%、80.85%。结论:总体上α-ASTREE型电子舌在天然药物四味辨识方面表现良好,同时在四味辨识中PCA-DA模型算法为最优,可以用于天然药物的四味辨识。Objective:In order to explore the feasibility of distinguishing natural medicines with different tastes based on electronic tongue and the difference of distinguishing effects of different types of electronic tongue.Methods:TS-5000 Z andα-ASTREE electronic tongues were used to detect 35 kinds of representative natural medicines and 12 kinds of food ingredients with four tastes of acid,bitter,sweet and salty,respectively,based on the extracted data of electronic tongue sensor and the results of traditional human taste panel method(THTPM),the qualitative discriminant model of four types of taste was established by using four algorithms,discriminant analysis(DA),least squares support vector machine(LS-SVM),principal component analysisdiscriminant analysis(PCA-DA),and partial least squares-discriminant analysis(PLS-DA).Results:Among them,the accuracy of the four algorithms(cross validation)based on TS-5000 Z electronic tongue data were 72.34%,72.34%,76.60%and 61.70%,respectively;the accuracy of the four algorithms(cross validation)based onα-ASTREE electronic tongue data were 91.49%,89.36%,93.62%and 80.85%,respectively.Conclusion:In general,theα-ASTREE electronic tongue performs well in the identification of four taste of natural medicines,and the PCA-DA model algorithm is optimal in the four-taste identification,which can be used for the four-taste identification of natural medicines.

关 键 词：天然药物 电子舌 经典人群口感评价 四味 定性判别 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] R2-03[自动化与计算机技术—控制科学与工程]