西洋参原材料近红外光谱精选研究

为确保药物成品质量的稳定均一,作为原材料的西洋参应保证纯正。采用近红外积分球漫反射光谱,首先对西洋参、人参原样光谱运用移动窗口偏最小二乘法(moving window partial least-squares regression,MWPLS)选择特征波段,建立最小二乘支持向量机(least-squares support vector machine,LS-SVM)种类鉴别模型挑选出西洋参样品,判别正确率为100%。然后将样品粉碎,人参以5%、10%、15%、20%掺入西洋参粉末,为增加样品代表性借助线性加和模拟光谱手段于4 000～7 000 cm-1建立主成分判别分析(principal component analysis discriminant analysis,PCA-DA)掺杂判别模型,预测判别正确率由97.2%上升至100%,区分出少量人参掺杂的西洋参样品,最终精选出西洋参原材料纯品。

Research on sifting of American ginseng as raw material with near infrared spectroscopy

To maintain the stability of drug products,it is necessary to ensure that American ginseng is pure as raw material.The near infrared integrating sphere diffuse reflective spectra are used.Firstly,moving window partial least-squares regression(MWPLS) is used to pick out the characteristic wavebands in the spectra of intact American ginseng and Panax ginseng.Then,the least-squares support vector machine(LS-SVM) model is built in the appointed wavebands to differentiate American ginseng from Panax ginseng with 100% accurate rate.Afterwards,all the samples are comminuted,and American ginseng powders are mixed with 5%,10%,15% and 20% Panax ginseng powders,respectively.To improve the representativeness of the samples,a principal component analysis discriminant analysis(PCA-DA) model is built with linear additive analog spectrum method at 4 000~7 000 cm-1;and the adulterant American ginseng is distinguished and the accurate rate is improved from 97.2% to 100%.As a result,the samples with a little Panax ginseng mixed are distinguished,and the pure American ginseng is sifted as raw material.