Abstract: The discrimination of kudzu starch with adulterated sweet potato starch and potato starch was conducted by mid-infrared spectroscopy (400－4000 cm-1) and principal component analysis (PCA). An equation was created through fisher linear discriminant analysis, and its reliability was verified by classifying training samples and testing samples. The results showed that the two dimensional distribution of principal component of samples belonged to different locations, and there was no overlap between kudzu starch location and adulterated samples. The classification of training samples and testing samples showed that the self-recognition rates of kudzu starch adulterated with sweet potato starch and potato starch were 93.3% and 100%, respectively. Meanwhile, the interactive correct recognition rates of kudzu starch adulterated with sweet potato starch and potato starch were 86.7% and 92.0%, respectively. So this method has good discrimination ability, and may offer a new approach for the rapid discrimination of kudzu starch adulterated with sweet potato starch and potato starch.

Key words: kudzu starch, Fourier transform infrared spectroscopy, principal component analysis, discriminant analysis