基于条件正定核的SVM人脸识别

Human face recognition based on conditionally positive definite kernel SVM

为提高人脸识别分类器的能力,采用了一种改进的可用于核学习方法的核函数—条件正定核函数。条件正定核函数一般不满足Mercer条件,但可以在核空间中计算样本间的距离,突出样本间的特征差异。对ORL、YALE、ESSEX三个标准人脸数据库进行仿真实验,结果表明基于条件正定核的SVM人脸识别算法在训练时间没有降低的情况下,与其他核函数法相比识别率有较大提高,并且当类别数增加时算法表现出较强的鲁棒性。

In order to improve the performance of classifier nel function named conditionally positive definite kernel.This for the face recognition, this paper introduces an improved kerkernel does not satisfy the well known Mercer condition,but it focuses on computing the distance between samples in the kernel space and finding the feature dissimilarity.Extensive face recognition experiments are based on three standard human face databases,namely ORL,YALE and ESSEX.The experimental results show that the proposed technique has higher recognition precision than other kemel function without more time on training.Furthermore,this method performs better robustness when the number of classes is adding.