Abstract: Near infrared diffuse reflectance spectroscopy combined with partial least square (NIRSPLS) was applied to analyzing the isoniazid in isoniazid tablets quantit atively and establishing quantitative mathematical calibration models of NIRS. In these models, the root mean square error of crosscalibration (RMSECV) was 0.00632.The root mean square error of prediction (RMSEP) was 0.006　03. The regression coefficient (R) was 0.99456and the average recovery of isoniazid was99.772%. In reappearance experiments, the relative standard deviation (RSD) was 0.526%.These results demonstrate that this method is of precise validation. Moreover, it also has the merits of convenience, nondestruction, no polluti on, good reappearance and is available for online detection.

Key words: near infrared spectroscopy, partial least square, isoniazid, nondest ructive analysis