摘要
为了解决流形学习不能充分利用样本类别信息的问题,提出了一种基于划分的有监督局部切空间排列算法,并将其应用于人脸识别。新算法采用基于动态粒子群算法的有监督的K均-值聚类算法确定样本的聚类中心,将样本划分为有重叠的块,新算法在利用数据类别信息的同时保持了流形的局部几何结构,提高了流形学习对图像的识别能力,能更好地适用于人脸识别。通过在ORL数据库上与其他流形方法比较,验证了新算法的有效性。
Taking into account that manifold learning couldn't make full use of sample category information,this paper presented a partitional supervised local tangent space alignment algorithm for face recognition.The new algorithm used supervised K-means clustering algorithm which based on dynamic management of particles to find clustering centers of data sample,and divided sample into overlapping parts.The new algorithm not only used the sample category information,but also kept the manifold local geometric structures.improved the identification ability for images of manifold learning.The new algorithm could do better in face recognition.Comparing with other manifold learning methods with ORL database show that the new algorithm is quite effective.
出处
《计算机应用研究》
CSCD
北大核心
2011年第6期2369-2371,共3页
Application Research of Computers
关键词
局部切空间排列法
有监督的K-均值聚类算法
动态粒子群算法
流形学习
人脸识别
local tangent space alignment
supervised K-means clustering algorithm
dynamic particle swarm optimization
manifold learning
face recognition