In this paper, a new method, feature extraction algorithm, for identification of zymic Pu-er tea based on Fourier Transform Infrared (FTIR) spectroscopy is proposed. Infrared spectra of 19 kinds of tea (including 6 kinds of zymic Pu-er tea) are recorded using Fourier Infrared Spectrometer. Six typical kinds of tea (including 4 kinds of zymic Pu-er tea) are chosen to establish training set, and the others (including 2 kinds of zymic Pu-er tea) are left to establish predicting set, which is used to check the accuracy of the identification by projecting. As a result, Puer characteristic basis and a stable linear arbiter can be obtained after the training set is dealt with feature extraction, and also the projecting result shows a clear identification: the zymic Pu-er teas in the predicting set are located in the same side of the linear arbiter as that in training set, meanwhile the others in the other side, which indicates that the linear arbiter is capable to identify zymic Pu-er tea.