For the problem that the feature extraction speed of palm vein recognition mothed based on local directional pattern (LDP) is slowly, a palm vein recognition method based on improved LDP is proposed. On the basis of keeping the Kirsch template of original LDP algorithm, the proposed method can code the edge response symbols in eight directions directly, which avoids sorting and reduces the time of feature extraction. It is called fast local directional pattern (FLDP). The proposed approach is tested on the palm vein datebases of Hong Kong Polytechnic University and the self-built palm vein datebases, the palm vein image is divided into some distributed blocks evenly, where the feature vectors of each sub block are obtained by FLDP algorithm and then the total characteristic fusion model is obtained, the chi-square distance is used for matching identification. Experimental results show that the proposed algorithm execution time is 17.75 ms and 19.98 ms respectively, whose performance obviously surpasses other typical algorithms, at the same time, the recognition rate can reach 99.8916% and 99.3665% . The proposed method has practical application value.