Abstract: Finger-back articular skin texture recognition is presented as a novel biometrics pattern. The hand back image is captured by the device made by ourselves. The articular skin texture is segmented and located by utilizing Canny operator and moving window. In the stage of recognition, two kinds of recognizing methods are given and compared: one is based on the correlation classifier and the other based on the complex Gabor wavelet transform. We judge the result of recognition using Hamming distance between two feature codes. The experiment results show that finger-back articular skin texture has such high uniqueness that can be used to authenticate human. The correct identification rate of the method based on correlation classifier is 98.04% when FAR equals FRR; the correct identification rate of the method based on Gabor wavelet transform is 94.61% when FAR equals FRR. But the later is faster than the former in the matching.