基于判别性降维的字典学习在人脸识别的应用

Application of Dictionary Learning Based on Discriminative Dimensionality Reduction in Face Recognition

针对目前大多数人脸识别方法只能单独实施降维或者字典学习而不能完全利用训练样本判别信息的问题,提出了基于判别性降维的字典学习方法,通过联合降维与字典学习使得投影矩阵和字典更好地相互拟合,从而可以获得更高效的人脸分类系统。所提方法的有效性在AR及MPIE两大通用人脸数据库上得到了验证,实验结果表明,相比于几种先进的线性表示方法,所提算法取得了更高的识别率,特别当训练样本数很少的时候,识别效果更佳。

For the issue that most face recognition methods can not use discriminative information of samples due to they only carry out dimensionality reduction or dictionary learning, dictionary learning method based on discriminative dimensionality reduction is proposed. Projection matrix and the dic- tionary can match with each other by jointing dimension reduction and dictionary learning, which will help to get more efficient face classification system. The effectiveness of proposed method is verified on AR and MPIE face databases. Experimental results show that proposed method gets higher recognition accuracy comparing with other latest linear represent approaches. Specially, its efficiency is better when training samples are not enough.