基于机器嗅觉结合BP神经网络的砂仁气味鉴别方法  

Odor Identification of Aromi Fructus Based on Machine Olfactory Combined with BP Neural Network

作　　者：朱志均 周华英 罗坤豪 吴恺熹 Zhu Zhijun;Zhou Huaying;Luo Kunhao;Wu Kaixi(College of Medical Information Engineering,Guangdong Pharmaceutical Universit)

机构地区：[1]广东药科大学医药信息工程学院

出　　处：《自动化与信息工程》2018年第4期45-48,共4页Automation & Information Engineering

基　　金：国家自然科学基金(61571140);全国大学生创新创业项目(201610573009)

摘　　要：检测和分析4种不同产地砂仁气味信息,探索基于机器嗅觉技术结合BP神经网络的气味鉴别方法。利用电子鼻PEN3采集4种不同产地砂仁的气味信息,构建一个包含三层结构的BP神经网络模型,通过对非线性气味数据进行反复学习和训练,确定各产地砂仁气味特征模型,采用MATLAB 2013实现鉴别仿真。结果表明:不同产地砂仁气味特征信息整体相貌是相似的,但各自的特征指纹峰值有明显差异;建立的BP神经网络模型对不同产地砂仁的训练样本和待测样本均可实现100%的准确鉴别。该方法稳定、可靠、简便,可为中药材质量控制和真伪鉴别提供参考。Four groups of aroma information of Aromi Fructus from different habitats were detected and analyzed, and a new identification method based on machine olfactory combined with BP neural network was put forward to identify Aromi Fructus from different habitats. Odor information of four groups of Aromi Fructus coming from different habitats was collected by an electronic nose （PEN3）, and a BP neural network model with three layers was constructed. Through repeated learning and training, models of odor characteristics of Aromi Fruetus were determined and the simulation for identification was realized based on MATLAB 2013. The experimental results showed that odor feature information of Aurantii Fructus from different growing areas were similar, but there were significant differences in peaks＇ value from each other. The BP neural network model was obtained and the accuracy rates for training set and testing set were all 100%. The method is simple, nondestructive and accurate, providing a reference for the quality control of Chinese herbal medicine.

关 键 词：机器嗅觉 BP神经网络 砂仁 气味鉴别 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] R282.5[自动化与计算机技术—控制科学与工程]