There are significant differences in taste and nutritional value among different varieties and producing areas of millet. Therefore, it is of reference significance for consumers to distinguish different kinds of millet. In this paper, Fourier transforms infrared (FT-IR) spectroscopy, two-dimensional correlation infrared (2D-IR) spectroscopy combined with curve fitting, principal component analysis (PCA) was used to distinguish varieties and origins of millet. The results showed that the original spectra of millet were similar, which were mainly composed of carbohydrates, proteins and lipids. The obvious differences in intensity were observed near 3 012, 2 962, 2 928, 2 856, 1 748 and 1 548 cm^-1 in SD-IR. The numbers, positions and intensities of auto-peaks and cross-peaks were different in the range of 1200~860 and 1700~1180 cm^-1. The curve fitting results showed that the ratio of the sub-peak areas of millet in the range of 1 700~1 600 cm^-1 was different, which indicated that the protein content of millet was different among different varieties, to realize the identification of millet varieties. The range of 1 800~800 cm^-1 in the derivative spectra was used for PCA analysis. The results showed that the cumulative contribution rate of the first three principal components was 97%, and millet from different producing areas was correctly classified. The study demonstrates that IR combined with statistical analysis methods could be effectively used to identify and analyze varieties and producing areas of millet.