摘要
针对主成分分析(PCA)算法对数据进行向量化,破坏初始数据的局部结构信息的缺点,提出了将局部线性嵌入(LLE)与PCA相结合的人脸识别算法。先采用LLE提取的初始数据保留了人脸局部结构信息的低维特征,再利用PCA计算低维数据的主要成分,最后根据各人脸的主要成分之间的欧式距离判断是否匹配。对比实验表明,该算法在明显提升算法效率的同时,保证了较高的识别率。
Focusing on the disadvantage that the principal component analysis (PCA) algorithm destroy the primary data of local structural information when it vectors the data, a face recognition algorithm that combines the locally linear embedding (LLE) with PCA was proposed. First, the low-dimensional features from the initial data which preserving the local structure information of face image was extracted by LLE. Secondly, the main components of the low-dimensional data with PCA were calculated. At last, the main components were judged whether they are matching or not according to their Euclidean distance. Comparative experimental result shows that this algorithm keeps a high recognition rate when improving the arithmetic efficiency.
出处
《重庆邮电大学学报(自然科学版)》
北大核心
2009年第1期92-94,114,共4页
Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)
基金
国家863计划项目(2005AA122310)
重庆市教委项目(KJ070511)
教育部新世纪优秀人才支持计划
重邮博士启动基金(A2007-60)
关键词
LLE
PCA
人脸识别
识别率
locally linear embedding (LLE)
principal component analysis (PCA)
face recognition
recognition rate