摘要
提出统计不相关的核化图嵌入算法,为求解各种统计不相关的核化降维算法提供了一种统一方法。与已有核化降维算法相比,新的特征提取方法降低甚至消除了最佳鉴别矢量间的统计相关性,提高了识别率。通过在ORL,YALE和FERET人脸库上的实验结果表明,提出的具有统计不相关的核化图嵌入算法在识别率方面好于已有的核算法。另外,揭示了统计不相关的核化图嵌入与已有的核化图嵌入的内在关系。
An uncorrelated kernel extension of graph embedding which provides a unified method for computing all kinds of uncorrelated kernel dimensionafity reduction algorithms is proposed. Compared with kernel dimensionality reduction methods, the proposed method is better in terms of reducing or eliminating the statistical correlation between features and improving the recognition rate. The experimental results on ORL, YALE and FERET face databases show that the proposed nncorrelated kernel extension of graph embedding method is better than other methods in terms of recognition rate. Besides, the relation between uncorrelated kernel extension of graph embedding and kernel extension of graph embedding is revealed.
出处
《中国图象图形学报》
CSCD
北大核心
2011年第4期618-624,共7页
Journal of Image and Graphics
基金
国家高技术研究发展计划(863)项目(2006AA01Z119)
国家自然科学基金项目(60873151
60632050)
高等学校博士学科点专项科研基金项目(20060288013)
江苏省2010年度普通高校研究生科研创新计划项目(178)
关键词
最佳鉴别矢量
统计不相关
核化图嵌入
optimal discriminant vectors
statistically uncorrelation
kernel extension of graph embedding （KGE）
