新颖人脸检测的核方法

Kernel method for novelty face detection

核主成分分析(kernel PCA)是PCA的非线性扩展。该研究将kernel PCA应用于新颖人脸检测。作为训练数据的人脸图像被映射到高维特征空间。在特征空间中,kernel PCA抽取数据分布的主成分,构成主子空间。在该子空间中,通过优化方法寻找包含训练数据的最小超平面,作为新颖检测的决策界面。该新方法在ORL人脸图像库的数据集中进行了实验,测试精度较线性PCA方法有所提高。

Kernel Principal Component Analysis（kernel PCA） is a non-linear extension of PCA.This paper introduces and investigates the use of kernel PCA for novel face detection.The face images for training data are mapped into a high-dimensional feature space.In this feature space,the principal components of the data distribution are extracted using kernel PCA to construct the main subspace.Then in the subspace,the hypersphere containing almost training data with minimum radius is found as decision hypersphere for novel detection.The experiments are implemented for proposed method using the ORL face database.The results reveal that the detection precision has a certain improvement compared with linear PCA.