摘要
采用伏安型电子舌对8种低钠盐配方样品进行苦味特性评价研究。使用电子舌系统采集样品信号,利用小波分析进行数据压缩预处理,结合主成分分析、聚类分析对低钠盐配方进行区分分类,利用径向基神经网络建立低钠盐苦味预测模型。结果表明:小波压缩后剩余能量和置零系数比分别为99.23%,99.42%;主成分得分图上,低钠盐配方与苦味参比样品差异明显;聚类分析结果与主成分分析结果相一致,正确反映了样品之间的亲疏关系;径向基神经网络建立的苦味预测模型均方根误差为1.48%,预测结果与实际感官评价结果相吻合。该研究为低钠盐的呈味特性评价提供了一种新的方法和途径。
A voltammetric electronic tongue is used to evaluate the bitterness of eight kinds of low-sodium salt samples. The signal of sample is collected by voltammetric electronic tongue system. The data are pretreated by wavelet analysis. And then the formula of low-sodium salt is classified by principal component analysis and cluster analysis. The model for predicting the bitterness of low-sodium salt is established by RBF neural network. The results show that the retained energy and the number of zeros are 99.23% and 99.42% after wavelet compression. On the score chart of principal component, the low-sodium salt formula and the bitterness reference samples have obviously differences. The results of cluster analysis and principal component analysis are consistent, and correctly reflect the relationship of samples. The root mean square error （RMSE） of the bitterness prediction model based on RBF neural network is 1.48%. The prediction results are in agreement with the actual sensory evaluation results. This research has provided a new way for the evaluation of flavor characteristics of low-sodium salt.
出处
《中国调味品》
北大核心
2017年第9期109-112,115,共5页
China Condiment
基金
粮食信息处理与控制教育部重点实验室资助项目(KF11-2015-101)
关键词
低钠盐
伏安型电子舌
苦味
小波压缩
主成分分析
径向基神经网络
low-sodium salt
voltammetric electronic tongue
bitterness
wavelet compression
principal component analysis tRBF neural network
