A method for personal identification based on hand geometric feature according to the measurable characteristic of hand shape is presented. Firstly feature points are localized by using curvature, according to which the absolute length of some parts of hand can be calculated. Then an eigenvector composed of relative lengths is obtained. The relative lengths in the eigenvector are generated from the absolute lengths of some parts of hand. At last, personal identification is achieved by computing Euclidean distance between eigenvectors. In the experiment, the accurate recognition rate is 91.08% by using only five characters, and 93.27% when the character number used to identify is 11. The experimental results show that, personal identification can be achieved by using the relative length of hand geometric feature. It can reach a high recognition rate only by use a few characters using the method proposed, and the method is easy to realize with characteristic of high utility.