基于加权多核学习的FLDA在人脸识别中的应用

模式识别技术的应用及研究表明,单核学习在人脸识别中的应用已经很成熟,但是,在单核学习中分类的效果并不是很好。基于此,提出了一种多核构造方法,即基于加权多核学习的FLDA方法(WMKL-FLDA),通过一系列带有权值约束的基本内核线性组合构建内核,并且利用权值优化迭代方案对最大边缘准则(MMC)进行优化。在FERET及CMU PIE人脸数据库上的实验表明,与以往的单核FLDA方法相比,提出的多核学习方法不仅实现了更高的识别性能,在构造内核方面也放松了参数的选择要求。

Application of FLDA Based on Weighting Multiple Kernel Learning in Face Recognition

Recent applications and researches of pattern recognition show that single kernel has been maturely applied in face recognition. However, classification efficiency of single kernel is not excellent. To address this problem, a multiple kernel construction method named Fisher Linear Discriminative Analysis based on Weighting Multiple Kernel Learning (WMKL-FLDA) is proposed in this paper. The constructed kernel is a linear combination of several base kernels with a constraint on their weights. By maximizing the margin maximization criterion (MMC), it presents an iterative scheme for weight optimization. The experiments on the FERET and CMU PIE face databases show that proposed multiple kernel learning method achieves high recognition performance comparing with single-kernel-based FDA and the constructed kernel relaxes parameter selection for kernel-based FLDA to some extent.