摘要
根据光寻址电位传感器(LAPS)原理,提出一种结合主成分分析和反向传播(BP)神经网络识别溶液味觉的方法。对LAPS电子舌采集的味觉数据主成分进行提取,将该主成分作为BP神经网络的训练样本,通过训练和学习构建味觉数据与味觉类别之间的联系,用训练后的BP网络对溶液进行味觉识别。对浓度分别为20 ppm、100 ppm、300 ppm和500 ppm的酸、甜、苦、咸、鲜5种味觉溶液进行识别验证,准确率达96.6%,结果表明该方法能够在不同浓度下正确识别出溶液的味觉。
According to the Light Addressable Potentiometric Sensor（LAPS） principle, a taste recognition system is proposed based on algorithms of Principal Component Analysis（PCA） and Neural Network（NN）. Using PCA to extract principal components of the original data, and the principal components are used as the training samples, to find the intrinsic links between the taste data and the taste type through learning and training process. The Back Propagation（BP） network trained is used to recognize the taste. Tastes of acid, sweet, bitter, salty and umami under density of 20 ppm, 100 ppm, 300 ppm and 500 ppm are recognized respectively. The rate of correctness is up to 96.6%, which shows that the taste recognition system can be used to identify the taste of liquid when the density is changed.
出处
《计算机工程》
CAS
CSCD
2012年第13期26-29,共4页
Computer Engineering
基金
国家"863"计划基金资助项目(2009AA04Z214)
关键词
光寻址电位传感器电子舌
味觉识别
神经网络
反向传播算法
主成分分析
特征提取
Light Addressable Potentiometric Sensor（LAPS） electronic tongue
taste recognition
neural network
Back Propagation（BP） algorithm
Principal Component Analysis（PCA）
feature extraction
