基于无相关判别稀疏投影的人脸识别方法

Uncorrelated sparsity discriminant projections for face recognition

针对人脸识别问题,提出了一种无相关判别稀疏投影(uncorrelated discriminant sparse projections,UDSP)方法。该方法通过设计一个无相关约束基于差的最优化目标,同时增加判别信息到稀疏保持投影(sparsity preserving projections,SPP)的目标函数,不仅保持了SPP的稀疏重构结构关系,而且利用了全局判别结构。同时,通过强加合适约束使得提取的特征统计无相关。最后,在FERET人脸库上进行了实验,证明了该方法的有效性。

A method of uncorrelated discriminant sparse projections （UDSP） is proposed .Through the design of a new uncorrelated constraint difference-based optimal objective and the addition of the ob-jective function from the discriminant information to the sparsity preserving projections （SPP ） ,the UDSP method not only preserves the sparse reconstructive relationship of SPP but also utilizes the global discriminant structures .Meanwhile ,an appropriate constraint is imposed to make the extracted features statistically uncorrelated .Experiments on the FERET face database verify the effectiveness of the proposed UDSP method .