深度特征筛选及融合的人脸识别方法研究

Research on selection and fusion of deep features for face recognition

针对人脸识别问题,提出采用深度特征筛选及融合的方法。采用卷积神经网络(CNN)学习人脸图像的多层次深度特征。对于所有的深度特征矢量,使用斯皮尔曼等级相关系数筛选其中有效部分。基于支持向量机(SVM)对筛选得到的任一深度特征矢量进行分类决策,并基于线性加权融合对它们的结果进行融合,最终确定待识别样本的人脸类别。基于ORL和Yale-B数据集对提出方法进行基础性能测试、噪声干扰稳健性测试及遮挡识别性能测试,结果验证了所提方法的性能优势。

Aiming at the problem of face recognition,a method based on selection and fusion of deep features is proposed.Use convolutional neural network(CNN)to learn the multi-level deep features of face images.For all the deep feature vectors,the effective part is selected with the Spearman rank correlation coefficient.Based on the support vector machine(SVM),the classification decision is made on any of the selected depth feature vectors,and their results are fused based on linear weighted fusion,and finally,the face category of the sample to be recognized is determined.Based on the ORL and Yale-B data sets,the basic performance test and noise interference robustness test of the proposed method are performed.The results verify the performance advantages of the method.