基于KDDA的人脸识别研究

Face Recognition Based on Kernel Direct Discriminant Analysis Algorithms

由于PCA和LDA算法存在小样本问题(Smell Sample Size),结合D-LDA和Kernel,将线性不可分的低维空间映射到高维空间,并借助于"kernel技巧"克服了维度灾难问题,并且充分的利用曾经被抛弃的有用信息Null-Space。经过才ORL人脸库的实验表明,此方法比PCA,LDA提高了人脸识别的可分性,并有效地解决了小样本问题。

Such as PCA（Principle Component Analysis） and LDA （Linear Discriminant Analysis ） exists SSS（＂small sample size＂） problem, we combine D-LDA and Kernel to project the line-unseparable feature space to the height dimensional separable feature space, and to overcome curse of dimensionality through kernel trick. It also use the Null-Space that has been aborted by some other method. The new algorithm has been tested, in terms of classification error rate performance,on the multi-view ORL face database. Result indicates that the proposed methodology is able to improve the recognition rate and solve the small sample size problem （SSS）.