摘要
为了快速辨识不同口味的玉米汁饮料,确保同一种饮料质量的均一性,构建了包含12个传感器的味觉传感器阵列。使用代表酸、甜、苦、咸、鲜的呈味物质检测味觉传感器阵列对5种基本味觉辨识的能力。使用主成分分析和概率神经网络考察了该阵列对基本味觉的辨识效果,该阵列对基本味觉表现出良好的辨识能力。将该阵列应用于玉米汁饮料的分类辨识中,区分来自不同品牌的9种玉米汁。系统聚类分析表明了同一种玉米汁样本的味觉特征非常接近,可聚合为一类。通过主成分分析法实现数据降维,提取前3个主成分作为概率神经网络的输入神经元。试验结果表明:该味觉传感器阵列对不同种玉米汁饮料具有较好的辨识能力,辨识的正确率为95.06%。
In order to identify corn juices with different flavor quickly and evaluate the conformance of the same corn juices, a taste sensor array including 12 sensors was built. The taste sensor array was tested with sweet, salty, sour, bitter and umami tastes as the evaluation of its ability to distinguish 5 basic tastes. Principal component analysis and Probabilistic neural networks were used for analyzing the effect to distinguish basic tastes based on the sensor array. The array allowed a successful recognition of the basic tastes. The taste recognition capability was further tested in the identification of corn juices. A total of 9 commercial corn juices from different brands were analyzed. Cluster analysis showed that taste characteristics from the same corn juices were similar, and aggregated as a cluster. Dimensionality reduction was achieved by Principal component analysis. The previous three principal components were applied as inputs of probabilistic neural networks. The taste sensor array showed good identitication of corn juices with identification accuracy of 95.06%.
出处
《农业工程学报》
EI
CAS
CSCD
北大核心
2012年第24期265-271,共7页
Transactions of the Chinese Society of Agricultural Engineering
基金
国家自然科学基金资助项目(31271861)
国家"863"高技术研究发展计划资助项目(2008AA100802)
关键词
传感器阵列
主成分分析法
聚类分析
玉米汁
分类辨识
神经网络
sensor arrays, principal component analysis, cluster analysis, corn juices, identification, neural networks
作者简介
刘晶晶（1986-），女，吉林白山人，博士生，研究方向：农产品智能检测与评价研究。长春吉林大学生物与农业工程学院，130022。Email．．1iujj1986tiger@sina．com
通信作者：孙永海（1956-），男，吉林长春人，教授，博士，博士生导师，研究方向：农产品智能检测与评价研究。长春吉林大学生物与农业工程学院，130022。Email：sunyh@jlu．edu．ca
