基于维度加权马氏距离的视频人脸识别

The Research of video face recognition based on Dimension weighted Mahalanobis Distance

针对人脸识别中图像集中附加信息利用率较低的问题,设计了一种维度加权的马氏距离度量方法,首先利用大间距架构中数据之间的相似关系来对马氏距离度量进行学习,再对其进行维度加权并结合K近邻分类器完进行人脸分类。结果表明:该方法在You Tube明星数据集上进行视频人脸识别的效果优于其它方法。

In order to solve the problem of low utilization ratio of additional information in face recognition, a dimension weighted Ma- halanobis distance measurement method is proposed Firstly, the Mahalanobis distance metric is studied by using the similarity between the data in the large space structure. Then With Mahalanobis distance metric is weighted,the K nearest neighbor classifier is used to classify the face The results show that the proposed method is superior to other methods in video face recognition on YouTube data set.