The work aims to establish a fast and nondestructive analysis method for testing paper express document bags. Sixty-three express document bag samples were tested by Fourier transform infrared spectroscopy. The peak position of the absorption peaks in the infrared spectrum was analyzed, and the spectral data were dimensionally reduced and classified in combination with principal component analysis. Fisher discriminant analysis was used to analyze and verify the classification results of express document bags. At the same time, two classification models of multilayer perceptron neural network and radial basis function neural network were established for analysis and verification. The 63 samples of paper express document bags could be divided into four categories, and the Fisher classification model could achieve 100% accuracy; the classification accuracy of the multi-layer perceptron neural network model was 95.23%, and that of the radial basis function neural network model was 92.06%. Through comparison, it was found that Fischer discriminant could achieve a more effective classification of paper express document bags. This method is simple, rapid, less sample consumption and non-destructive to samples. It can provide a scientific basis for identification of physical evidence of express document bags.