Fourier transform infrared spectroscopy combined with discrete wavelet transform was used to discriminate 6 species of melon seeds. Spectra of 30 samples were obtained. The results showed that the infrared spectra of the samples were similar, while in the range of 1700～900 cm^-1, tiny differences in wave-numbers and absorption intensity of peaks were observed. The spectra in this range were selected to perform discrete wavelet transform, and the obvious differences were found in the second scale detail coefficients. The second scale coefficients were selected to perform the correlation analysis, principal component analysis (PCA) and hierarchical cluster analysis (HCA). The results showed that towel gourd seeds had very significant positive correlation with cushaw seeds with correlation coefficient 0.9279, while the towel gourd seeds and semen benincasae had minimum correlation coefficient. The contributing rates of the first two principal components in PCA had reached 95.50%, which can reflect the main information in the spectra. The accuracy of PCA and HCA were 100%. It demonstrates that Fourier Transform Infrared spectroscopy combined with discrete wavelet transform was a rapid and effective method for discriminating different species of melon seeds.