A discriminant analysis technique using wavelet transformation(WT)and influence matrixanalysis(CAIMAN)method is proposed for the near infrared(NIR)spectroscopy classifi-cation.In the proposed methodology,NIR spectra are decomposed by WT for data com-pression and a forward feature selection is further employed to extract the relevant informationfrom the wavelet coefficients,reducing both classification errors and model complexity.Adiscriminant-CAIMAN(D-CAIMAN)method is utilized to build the classification model inwavelet domain on the basis of reduced wavelet coefficients of spectral variables.NIR spectradata set of 265 salviae miltiorrhizae radia samples from 9 different geographical origins is usedas an example to test the classification performance of the algorithm.For a comparison,k-nearest neighbor(KNN),linear discriminant analysis(LDA)and quadratic discriminant analysis(QDA)methods are also employed.D-CAIMAN with wavelet-based feature selection(WD-CAIMAN)method shows the best performance,achieving the total classification rate of ioo%in both cross-validation set and prediction set.It is worth noting that the WD-CAIMANclassifier also shows improved sensitivity,selectivity and model interpretability in thecla.ssifications.
Homogeneity of powder blend is essential to obtain uniform contents for the tablets and capsules.Near-infrared(NIR)spectroscopy with fiber-optic probe was used as an on-line technique for monitoring the homogeneity of pharmaceutical blend during the blending process instead of the traditional techniques,such as high performance liquid chromatograph(HPLC)method.In this paper NIRS with a SabIR difuse reflectance fiber optic probe was used to monitor the blending process of coptis powder and lactose(excipient)with different contents,and further qualitative methods,like similarity,moving block of standard deviation and mean square were used for calculation purposes with the ollected spectra after the pretreatment of multiplicative signal correction(MSC)and second derivative.Correlation spectrum was used for the wavelength selection.Four different coptis were blended with lactose separately to validate the proposed method,and the blending process of "liu wei di huang"pill was also simulated in bottles to verify this method on multiple herbal blends.The overall results suggest that NIRS is a simple,efective and noninvasive technique can be sucssfuly applied to the determination of homogeneity in the herbal blend.