基于字典投影学习的人脸识别算法

针对字典学习l0或l1范数的稀疏约束导致训练和测试阶段较高的复杂性,提出用于人脸识别的字典投影学习算法.该算法合成和分析字典,达到信号表示和分类.实验结果表明,与传统的DL方法相比,所提出的DPL方法大大降低了训练和测试阶段的时间复杂度;与KNN算法相比,具有较高的识别精度和较好的稳定性.

Dictionary Projective Learning for Face Recognition

According to the sparse constraint with l0 or l1norm in dictionary learning,which results in the higher complexity during the training and testing process,the dictionary projective learning for face recognition was proposed by synthesizing and analyzing dictionary to represent and classify signals. The experimental results showed that the proposed dictionary project algorithms( DPL) reduced the complexity of training and testing compared traditional dictionary learning( DL),and had higher recognition accuracy and better stability with K-Nearest Neighbors( KNN).