Human authentication plays a crucial role in sensitive applications like ATM usage, entry into a secured area, attendance and many more. A novel human authentication system is proposed by extracting curvelet multiresolution features from the palm vein trait. The entire palm region is extracted by using an improved bounding rectangle strategy and is further enhanced using Difference of Gaussian (DoG) and Histogram Equalization (HE) methods in order to make the vein pattern, more prominent. Curvelet, a multiresolution transform which handles curve discontinuities well is applied with five scales and sixteen orientations over the enhanced palm vein region. Standard deviation and mean features are calculated from the obtained curvelet subbands. Two scores are computed from these individual features and finally fused using weighted sum rule. The experiments are conducted in publicly available CASIA and VERA palm vein databases which results with the recognition rate of 99.7% and 99.86%, respectively. The proposed system achieved the lowest equal error rate (EER) of 0.021% and 0.0207%, respectively, for CASIA and VERA palm vein database as compared with other state-of-the-art methods. The system performance measured in terms of computation time took a maximum of 0.09 s in identifying an individual.

人工身份验证在 ATM 使用、进入安全区域、出勤等敏感应用中起着至关重要的作用。通过从手掌静脉特征中提取曲波多分辨率特征，提出了一种新的人体认证系统。通过使用改进的边界矩形策略提取整个手掌区域，并使用高斯差分（DoG）和直方图均衡（HE）方法进一步增强，以使静脉图案更加突出。Curvelet，一种可以很好地处理曲线不连续性的多分辨率变换，在增强的手掌静脉区域上应用了五个尺度和十六个方向。从获得的曲波子带计算标准偏差和平均特征。根据这些单独的特征计算两个分数，最后使用加权求和规则进行融合。实验在公开可用的 CASIA 和 VERA 手掌静脉数据库中进行，结果识别率分别为 99.7% 和 99.86%。与其他最先进的方法相比，所提出的系统分别为 CASIA 和 VERA 手掌静脉数据库实现了 0.021% 和 0.0207% 的最低等错误率 (EER)。以计算时间衡量的系统性能在识别个人时最多需要 0.09 秒。