摘要
掌纹识别是生物特征识别技术的新热点,论文提出使用二维主成分分析算法(2D PCA)提取掌纹图像的统计特征,实验表明其泛化能力优于传统主成分分析算法(PCA).在此基础上,论文提出且定义了改进的二维主成分分析,并证明它在保持训练样本图像总体散度的同时更有效的提取样本特征.改进的算法在得到99.72%高识别率的同时,大幅降低了原算法的特征维数、识别计算的复杂度,使系统的实用性进一步提高.
Palmprint recognition is one of the most reliable and new technique of biometrics. This paper presents a new approach based on 2D PCA algorithm, which extracts the statistical feature of palmprint images. Experimental results illustrate its performance and generalization ability over PCA. Furthermore, we propose improved 2D PCA method, and prove that the algorithm can hold the global scatter of the training images while decorrelates rows as well as columns of them. In this way,we obtain high accuracy （99. 72 % ） while reducing the feature dimension efficiently, the responding time of recognition is only 0.03s, which enhances the practicability of the system.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2005年第10期1886-1889,共4页
Acta Electronica Sinica
