摘要
本文对大豆分离蛋白在风味蛋白酶五种不同酶解时间下制取的大豆肽溶液进行区分和苦味评价。采用法国ASTREE电子舌传感器采集大豆肽溶液的味觉信息,基于判别因子分析法(DFA)对大豆肽溶液进行定性分析;基于偏最小二乘法与RBF神经网络分别建立苦味得分的定量预测模型。结果显示DFA对不同大豆肽样品的区分效果好,区分指数DI=100,并能准确对未知样品进行呈味预测,识别率为100%;采用偏最小二乘法模型时的建模集和预测集的RMSE分别为2.47%、6.81%,采用RBF神经网络模型时的RMSE分别为0.81%、3.37%,表明采用RBF模型的预测效果比偏最小二乘法好。研究结果可为后续大豆肽产品的呈味特性研究提供一种新的方法途径。
In this paper,the discrimination and bitter flavor characteristics assessments of five different samples was studied. The taste information of soybean peptide solution was collected by the French ASTRESS electronic tongue sensor and the method of DFA was applied to do qualitative analysis of the solution. The relationship between the sensor response value and bitter taste score was analyzed and two qualitative prediction modes was established based respectively on the methods of RBF and partial least squares. The results showed that the method of DFA had a good discrimination of different soybean peptide samples with100 distinction index and accurately predicted the taste of unknown samples with 100% recognition rate. The RMSE based on partial least squares of modeling and prediction sets were 2.47% and 6.81% respectively while the model based on the methods of RBF were 0.81% and 3.37%. It showed that the model based on RBF did better in the prediction than the model based on partial least squares and the results might provide a way to the follow-up study of the flavor characteristics of soybean peptide product.
出处
《食品工业科技》
CAS
CSCD
北大核心
2016年第8期97-99,109,共4页
Science and Technology of Food Industry
