How to fast ID fake?

However, complicated operations and costly equipments of these technologies restricted their applicability,especially for the developing areas/ countries．In the contrary, spectroscopy has the natural advantage of convenient, cheap and fast ID for the prescreening purpose.

In this study, we establish two typies of simple and fast methods based on Raman spectroscopy and chemometrics to determine the counterfeit drugs without any sample pretreatment. One is the system of counterfeit drugs discriminant analysis，which included three analytical modules. The other is the method based on the Raman spectroscopy and discriminative model with small training set to determination of hypoglycemic drugs.

Part I The system of rapid detection of counterfeit drugs: The system is proposed according to the characteristics of Raman spectrums of active pharmaceutical ingredient and true drugs. The system consists of three modules．They are overall spectrum module, characteristic segment spectrum module and characteristic peaks module, which is criminate the possible counterfeit(s)from point, line and flat angles，respectively. The second module is the core of the system. This module applies the convolution function of convolution spectroscopy to Raman spectroscopy; one spectrum Can be converted into hundreds of thousands of convolution curve. According to the characteristic of the convolution curve，some characteristic segments of the convolution curve are screened to determine the counterfeit drugs. SO we call the module, convolution curve screening(CCS) method. The algorithms are exemplified by six kinds of hypoglycemic(glibenclarnide, gliclazid, pioglitazone，glipizide and Acarbose)tablets and 12 common pharmaceutical excipients. The total sensitivity，specificity，accuracy and efficiency reach 97.50％, 100.0％, 99.60％and 96.35％ respectively.

PartⅡ Determination of hypoglycemic drugs by Raman spectroscopy-discriminatlVe model with small training set. In order to discriminate hypoglycemic drugs by the Raman spectroscopy technology, six kinds of hypoglycemic tablets。including 1 06 samples，were gathered and analyzed using a portable Raman spectrometer. The sample data were pretreated with the methods of baseline correction, smoothing and vector normalization, followed by principal component analysis(PCA). The anterior 10 principal components were then used as the new variables, and analyzed by Fisher multi-types linear discriminate, Bayes multi, types stepwise discriminate, K-nearest neighbor(KNN)and radial basis function(RBF) neural network algorithm, respectively．The results demonstrated that each of the four methods provided rapid, non. destructive identification of hypoglycemic drugs. The most prominent among them is PCA combined with the method of Bayes multi_types stepwise discriminate analysis, which offers an effective approach to the rapid discrimination of different kinds of hypoglycemic tablets.