摘要
论文针对人脸识别中的特征选择和单一算法的局限性问题,提出了一种基于深度学习的局部方向人脸识别算法。首先通过定位和分块选取人脸局部敏感区域,然后依次利用LDP算法良好的局部特征提取能力对精选分块区域进行特征提取并最终连接成新的特征脸,然后利用深度学习网络DBN进行逐层贪婪训练,获得良好的网络参数,最后用训练好的网络对测试样本进行人脸测试分类。依次在ORL和MIT-CBCL人脸数据库上进行实验检测,实验结果表明,论文提出的算法与传统单一或融合算法具有更高的识别率,具有良好的局部性能和抗干扰性。
The paper proposes a local directional face recognition algorithm based on deep learning for feature selection in face recognition and the limitation of single algorithm. Firstly,the local sensitive area of the face is selected by positioning and seg. mentation. Then the LDP algorithm's good local feature extraction capability is used to perform feature extraction on the featured block region and finally connect it into a new feature face. Then a layer-by-layer greedy training is performed by using the deep learning network DBN to obtain good network parameters,and finally the trained network is used to test samples for face classifica. tion. Experiments are performed on ORL and MIT-CBCL face databases in sequence. The experimental results show that the pro. posed algorithm has a higher recognition rate than the traditional single or fusion algorithm and has good local performance and an. ti-interference performance.
作者
公维军
吴建军
李晓霞
李晓旭
GONG Weijun;WU Jianjun;LI Xiaoxia;LI Xiaoxu(School of Information Technology and Communication,Hexi University,Zhangye 734000;School of Computer and Communication,Lanzhou University of Technology,Lanzhou 730050)
出处
《计算机与数字工程》
2019年第5期1032-1036,1135,共6页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号:61563030)
河西学院青年教师科研基金项目(编号:QN2017014)资助
关键词
局部方向模式
局部敏感
深度学习
特征选择
人脸识别
local directional pattern
local sensitive
deep learning
feature selection
face recognition
