基于核Fisher鉴别分析的手指静脉识别

A finger vein recognition method which is based on kernel Fisher discriminant analysis

抽取最优鉴别特征是手指静脉识别中重要的一步,在提取小样本的高维手指静脉图像特征时,由于光照、温湿度、水平位移等因素的影响使得采集的静脉图像是非线性分布的,为此,提出了一种基于核Fisher鉴别分析(kernel Fisher discriminant analysis,KFDA)提取非线性特征的方法。该方法是利用一个核映射将原始输入空间变换到一个更低维的空间RN中,在此特征空间上用核类间散度阵和核类内散度阵作为Fisher线性判别准则(Fisher linear discriminant,FLD),来得到最佳非线性鉴别特征,根据此鉴别特征计算其相互间的欧式距离进行识别。实验结果表明,核Fisher方法与其他方法相比,具有较低的认假率(false accept rate,FAR)和较快的识别速度。

Extracting the optimal discriminating characteristics is an important step of finger vein recognition.In extracting the image feature of high dimensional finger vein of small samples,the influence of the light,temperature,humidity and horizontal displacement makes the acquisition of the vein image＇s distribution nonlinear.Therefore,a method based on nuclear Fisher＇s（Kernel Fisher discriminant analysis,KFDA）differential analysis to extract the nonlinear characteristics is put forward.This method uses a nuclear mapping to change the original input space to a more low-dimensional space,using the dispersion matrix between the nuclei as the Fisher（Fisher linear discriminant,FLD）standard to get the best nonlinear differential features,Finally,according to the identification of characteristics the determination is done by using the calculation of the Euclidean distance between each other.The experimental results show that the method of nuclear has the lower recognize false rate FAR（false accept rate）and faster speed of recognition than other methods.