New approach for discriminant analysis of adulterated milk is proposed based on combining hetero-spectral two-dimensional (2D) near-infrared (NIR) and mid-infrared (IR) correlation spectroscopy along with multi-way partial least squares discriminant analysis (NPLS-DA).Firstly,36 pure milk samples were collected and 36 adulterated milk with starch samples (0.01 to 1 g·L-1) were prepared by adding appropriate mass of starch into pure milk.Then,one-dimensional NIR transmittance spectra and IR attenuated total reflection spectra of pure milk and adulterated milk with starch were measured at room temperature.And the synchronous 2D NIR-IR (4 200~4 800 vs.900~1 700 cm-1) correlation spectra of all samples were calculated.Due to the trace of adulterants,the synchronous 2D IR-NIR correlation spectral differences between adulterated milk with starch and pure milk are very subtle.Consequently,it was impossible to directly distinguish whether the sample was pure milk or adulterated milk.Finally,2D IR-NIR correlation spectra were to build a discriminant model to classify adulterated milk and pure milk.The classification accuracy rates of samples in calibration set and in prediction set were 95.8% and 100% respectively.Also,the NPLS-DA models were built based on 2D NIR and 2D IR correlation spectra,respectively.The classification accuracy rates of samples in prediction set were 95.8%.Comparison results showed that the NPLS-DA model could provide better results using 2D NIR-IR correlation spectra than using 2D NIR,and 2D IR correlation spectra.The proposed method can not only effectively extract the feature information of adulterants in milk,but also explores a new perspective method for detection of adulterated food.