The finger width is one of the effective features in the hand shape recognition,and the selected measuring points effect on the accuracy and recognition time.This paper researches the relationship among the selected points,recognition rates and matching time to estimate the selected points in different fingers according to different required recognition rates.Obtained results prove that the recognition rate is more than 90% under the conditions of accurate localization and selected points more than 3 in each finger.An automatic recognition algorithm is proposed,in which the curvature is used to localize feature points,then it used to calculate the absolute widths of fingers.After the eigenvectors are formed by the relative widths between two absolute widths,the Euripides distance between the two eigenvectors is computed to implement the feature recognition.Experiments show that feature values can be acquired to distinguish the human's identity when 6 points of each finger are selected,and the recognition rate can reach 94.03%.Experiment results present the two relationship curves of the number of points vs the matching time and the number of points vs the recognition rate.