基于电子舌和模式识别的中成药品辨识方法研究

发布时间：2018-06-12 17:48

本文选题：虚拟仪器 + 电子舌　；参考：《电子测量与仪器学报》2017年07期

【摘要】：为了实现不同中成药品的快速区分辨识,采用基于虚拟仪器技术的伏安电子舌系统对治疗感冒病症的4种不同中成药品进行了检测分析。分别采用特征点提取(FPE)法和离散小波变换(DWT)法对电子舌输出信号进行预处理,以样本点的聚集性和分类效果为依据,确定较佳的特征提取方法为以db4为母小波进行的8层离散小波变换。在此基础上,分别采用主成分分析法(PCA)、聚类分析法(CA)和BP神经网络(BPNN)对不同中成药品进行区分辨识。结果表明,PCA结果中PC1和PC2累计贡献率为95.6%,除羚羊感冒片和银翘解毒片有重叠趋势外,其余各类得到有效区分;CA能够有效地观察出4种中成药品之间的差异程度,但4种药品最终被分成两类,区分效果较差;非线性分类模型BPNN对不同中成药品区分效果较好。通过优化实验,分别确定了模型的训练算法、激活函数和隐含层节点数目等参数,测试集验证表明,BPNN模型对4种中成药品的分类正确率达到100%。本研究结果可为中成药品的非感官质量评价和快速辨识研究提供技术参考。
[Abstract]:In order to realize the rapid identification of different proprietary Chinese medicines, four kinds of Chinese patent medicines for the treatment of colds were detected and analyzed by using the voltammetry electronic tongue system based on virtual instrument technology. The feature point extraction (FPE) method and discrete wavelet transform (DWT) method are used to preprocess the output signals of electronic tongue, which are based on the aggregation of sample points and the classification effect. The best feature extraction method is 8-layer discrete wavelet transform based on db4 wavelet. On this basis, the principal component analysis (PCA), cluster analysis (CAA) and BP neural network (BPNN) were used to distinguish and identify different Chinese patent medicines. The results showed that the cumulative contribution rate of PC1 and PC2 in PCA was 95.6. except for antelope Ganmao tablets and Yinqiao jiedu tablets, the differences between four kinds of proprietary medicines could be observed effectively. But the four drugs were divided into two categories, and the nonlinear classification model (BPNN) was better than BPNN in differentiating different Chinese patent medicines. Through optimization experiments, the training algorithm, activation function and number of hidden layer nodes are determined, respectively. The test results show that the classification accuracy of BPNN model for four kinds of Chinese patent medicines is 100%. The results of this study can provide technical reference for non-sensory quality evaluation and fast identification of Chinese patent medicine.
【作者单位】：山东理工大学计算机科学与技术学院;山东淄博市中西医结合医院;山东淄博昌国医院;
【基金】：山东省自然科学基金(2015CM016)资助项目
【分类号】：R286;TP212.9;TP391.4

【相似文献】