Abstract: One of the most important part of the palmprint recognition is feature extraction. In this paper, we propose a new method to extract the palmprint features by the real discrete Gabor transform, which converts the spatial domain palmprint to spatial-frequency domain. We take the palmprint energy distribution in spatial-frequency domain as palmprint features. Then we make a feature match based on energy distribution features by using Euclidean distance and Support vector machine respectively. The experiment results prove that this method is robust to the palmprint with small move, small angle rotation, small flex and achieves high recognition rate.

Key words: Palmprint, Discrete gabor transform, support vector machine

摘要： 掌纹识别作为一种重要的生物特征识别方法，其中的一个重要环节就是掌纹特征的提取，本文提出了一种基于实数形式离散Gabor变换的掌纹特征提取方法，将空域的掌纹图像变换到联合（时间）空间频率域并将其联合（时间）空间频率域的能量分布作为掌纹的特征，以此为基础分别使用欧式距离和支持向量机进行了不同掌纹的匹配识别。实验结果表明，该算法对掌纹图像小的平移、小角度的旋转和小的手掌伸缩具有鲁棒性，并且获得了较高的识别率。

关键词: 掌纹, 离散Gabor变换, 支持向量机