摘要
选用10种不同口味的醋豆产品,在4种不同的稀释浓度下,用AstreeII电子舌采集数据,对采集到的味觉信号数据进行分析和处理,并建立费歇尔(Fisher)多级判别模型、三层BP神经网络模型对醋豆的类型进行判别,研究发现在醋豆溶液的稀释倍数为250×10倍时,2个模型都能达到较好的预测效果,Fisher多级判别的正确识别率达到95.70%,交互验证的正确识别率达到93.50%,三层BP神经网络的训练集正确识别率达到85.87%,测试集正确识别率达到78.26%。
The test in this paper makes use of Astree II electronic tongue to distinguish 10 kinds of vinegarbeans. The experiment was carried out in four different levels of vinegar-bean solutions. The data of taste signals was analyzed and processed, and then the Fisher discriminatory analysis model and three-layer back propagation artificial neural network model were developed. The results show that the prediction of 2 models bring about high effect and attain the accuracy requirements when the dilution rate is 250×10. The Fisher discriminatory analysis model＇s accurate identification rate is 95.70%, the cross-validation accurate identification rate is 93.50%, BP-ANN model＇s testing set accurate identification rate is 85.87%, training set accurate identification rate is 78.26%.
出处
《食品科技》
CAS
北大核心
2009年第9期290-294,共5页
Food Science and Technology
关键词
电子舌技术
醋豆
费歇尔判别分析
人工神经网络
electronic tongue technology
vinegar-bean
Fisher discriminant analysis
artificial neural network
