详细信息
文献类型:期刊文献
中文题名:基于流形学习算法的人脸识别研究
英文题名:Based on the Manifold Learning Algorithm of Facerecognition
作者:高峥[1];杜川[2]
第一作者:高峥
机构:[1]新乡学院现代教育中心;[2]新乡学院机电工程学院
第一机构:新乡学院
年份:2011
卷号:19
期号:4
起止页码:24-25
中文期刊名:河南机电高等专科学校学报
外文期刊名:Journal of Henan Mechanical and Electrical Engineering College
语种:中文
中文关键词:流形学习;人脸识别;监督的保持邻域嵌入;度量优化的保持邻域嵌入
外文关键词:face recognition; manifold learning; supervised neighborhood preserving embedding; metric-optimized neighborhood preserving embedding
摘要:为了有效地实现对人脸的识别,先用线性判断分析(LDA)方法将原始的人脸数据降维,利用降维后的数据选取点的k近邻,进而提出度量优化的保持邻域嵌入算法(MONPE)。MONPE算法:一方面,通过LDA降低原始数据的维数,使得欧氏度量的应用成为合理。另一方面,通过LDA拉近了类内点的距离,拉大了类间点的距离,使得某个采样的近邻点属于同一类的概率更大。
For achieving the recognition of face,firstly,with a linear analysis(LDA) method will be the original face data dimensionality reduction using low-dimensional data selection of nearest neighbors,and then put forward the measure to keep the neighbor embedding algorithm optimization(MONPE).MONPE algorithm: on the one hand,through the LDA to reduce the original data dimension,the Euclidean metric of the application as reasonable.On the other hand,through the LDA close within class point distance,pull big class point distance,makes a sampling of the neighbor points belonging to the same class a greater probability.
参考文献:
正在载入数据...