The adulteration of traditional Chinese medicine not only reduces the curative effect but also seriously harms the health of body.In this paper,the near infrared spectroscopy technique was used to qua ntitatively analyze the content of talc powder in angelica dahurica.The NIR diffuse reflectance spectra of 181samples mixed with talc powder and 7samples mixed with talc powder and flour were measured by a self-developed near infrared spectrometer w hich works in the range of 900－1700nm with a resolution of 4nm.Combining with smoothing and first derivative method the spe ctral data were first to be preprocessed.Then,the performance of models based on different algorithms was studied.A new optimized quantitative method MW-GA-LS-SVM was proposed.Using the moving window method (MW) combined with genetic algorithm (G A) to select characteristic wavelength,then build the model by least square support vector machine (LS-SVM) method.Result showed that MW-GA can effectively select wavelength and improve the predictive accuracy.And the model was validated with multi-component samples which were mixed with talc powder and flour.Compared with full-spectrum-PLS,MW-GA-PLS and f ull-spectrum-LS-SVM,MW-GA-LS-SVM had the best performance with the R²=0.995,RMS EC=0.42%,RMSEP=0.48%.The result showed that the model established by MW-GA-LS-SVM has high accuracy and provides a method for the application of near infrared spectroscopy in the rapid quantitative analysis of Chinese medicine ingredients.