基于深度自编码器的单样本人脸识别

发布时间：2018-03-22 16:44

本文选题：单样本人脸识别　切入点：深度自编码器　出处：《模式识别与人工智能》2017年04期 　论文类型：期刊论文

【摘要】：由于每个目标仅有一幅已知样本,无法描述目标的类内变化,诸多人脸识别算法在解决单样本人脸识别问题时识别性能较低.因此文中提出基于深度自编码器的单样本人脸识别算法.算法首先采用所有已知样本训练深度自编码器,得到广义深度自编码器,然后使用每个单样本目标的单个样本微调广义深度自编码器,得到特定类别的深度自编码器.识别时,将识别图像输入每个特定类别的深度自编码器,得到包含与测试图像相同类内变化的该类别的重构图像,使用重构图像训练Softmax回归模型,分类测试图像.在公共测试库上进行测试,并与其它算法在相同环境下进行对比,结果表明文中算法在获得更优识别率的同时,识别一幅图像所需平均时间更少.
[Abstract]:Because each target has only one known sample, it is not possible to describe the in-class variation of the target, Many face recognition algorithms have low recognition performance in solving the problem of single sample face recognition. Therefore, a single sample face recognition algorithm based on depth self-encoder is proposed in this paper. Firstly, all known samples are used to train depth self-encoder. The generalized depth self-encoder is obtained, and then the generalized depth self-encoder is obtained by fine-tuning the generalized depth self-coder with a single sample target, and the recognition image is input into the depth self-coder of each specific class. The reconstructed images containing the same intra-class changes as the test images are obtained. The reconstructed images are used to train the Softmax regression model and classify the test images. The images are tested on the common test library and compared with other algorithms in the same environment. The results show that the algorithm achieves better recognition rate and the average time required to recognize an image is less.
【作者单位】：中国人民解放军信息工程大学信息系统工程学院;郑州轻工业学院计算机与通信工程学院;
【基金】：河南省教育厅科学技术研究重点项目(No.12A510027)资助~~
【分类号】：TP391.41

【相似文献】