KPCA FACE RECOGNlTION METHOD
A. Overview
The nuclear field of pattern recognition method is the
rapid development of a new direction; a series of advanced
nonlinear data processing technology, the common feature is
that these data processing methods applied to nuclear
mapping. The most commonly used nuclear methods include:
support vector machine, support vector regression, kernel
principal component analysis, nuclear Fisher criterion, the
classification based on nuclear and other kernel-based
projection pursuit. Its main idea was originally proposed by
V. Vapnik [1, 2] who proposed and used in support vector
machine (SVM). Scholkopf et al [3, 4, 5, 6] implemented
that nuclear method is applied to feature extraction, and
proposed kernel principal component analysis (KPCA),
experimental results show that KPCA can not only extract
nonlinear features, but also better recognition results.
From the specific operation point of view, nuclear
nonlinear mapping method first used the raw data from the
data space is mapped to high dimensional feature space, then
in the feature space corresponding to linear operation.
Because we use the nonlinear mapping, this nonlinear
mapping is often very complex, thereby greatly enhancing
the ability of nonlinear data processing.