一种通过卷积─池化提升SVM人脸识别率的研究

楚雄师范学院学报 ›› 2020, Vol. 35 ›› Issue (3): 106-114.

一种通过卷积─池化提升SVM人脸识别率的研究

叶晓波¹, 秦海菲², 吕永林³

1.楚雄师范学院网络与信息系统研究所,云南楚雄 675000;
2.楚雄师范学院信息科学与技术学院,云南楚雄 675000;
3.楚雄师范学院经济与管理学院,云南楚雄 675000

收稿日期:2019-12-17 出版日期:2020-05-20 发布日期:2020-12-28
作者简介:叶晓波(1975-),男,硕士,楚雄师范学院网络与信息系统研究所副教授,研究方向为模式识别、现代教育技术。E-mail:yncxyxb@hotmail.com

Research on Improving SVM Face Recognition Rate by Convolution - Pooling

YE Xiaobo¹, QIN Haifei², LV Yonglin³

1. Institute of Network & Information Systems,Chuxiong Normal University,Chuxiong,Yunnan Province 675000;
2. School of Information Sciences & Technology,Chuxiong Normal University,Chuxiong,Yunnan Province 675000;
3. School of Economics & Management,Chuxiong Normal University,Chuxiong,Yunnan Province 675000

Received:2019-12-17 Online:2020-05-20 Published:2020-12-28

摘要/Abstract

摘要： 以剑桥大学计算机实验室的ORL Faces数据库作为实验数据,通过卷积神经网络中的“卷积-池化”层对实验数据进行处理,选择LIBSVM集成软件为工具,对原始数据和经“卷积-池化”处理后的数据进行了分类识别研究,SVM参数选用C-SVC模型、nu-SVC模型与线性核函数、多项式核函数、径向基核函数、Sigmoid核函数进行组合。实验结论:增加训练数据可提高人脸识别率,卷积-池化处理可实现数据降维,“高斯平滑卷积核卷积-池化”处理可提高SVM人脸识别率,SVM在人脸识别中更适合选用C-SVC模型+线性核函数。

关键词: 人脸识别, 卷积, 池化, SVM(支持向量机)

Abstract: Taking the ORL Faces database of the computer laboratory of Cambridge University as the experimental data,the experimental data are processed through the "convolution-pooling" layer in the convolution neural network,and the LIBSVM integration software is selected as the tool to classify and identify the original data and the data processed by "convolution-pooling".SVM parameters are combined with C-SVC model,nu-SVC model and linear kernel function,polynomial kernel function,radial basis kernel function and Sigmoid kernel function.The convolution-pooling processing can achieve data dimensionality reduction and "gaussian smooth convolution kernel convolution-pooling" processing can improve SVM face recognition rate.SVM is more suitable to choose C-SVC model linear kernel function in face recognition.

Key words: face recognition, convolution, pool, SVM

中图分类号:

TP18

叶晓波, 秦海菲, 吕永林. 一种通过卷积─池化提升SVM人脸识别率的研究[J]. 楚雄师范学院学报, 2020, 35(3): 106-114.

YE Xiaobo, QIN Haifei, LV Yonglin. Research on Improving SVM Face Recognition Rate by Convolution - Pooling[J]. Journal of Chuxiong Normal University, 2020, 35(3): 106-114.

参考文献

[1] Sergios T,Konstantinos K.Pattern Recognition,Fourth Edition[M].Academic Press;4 edition (November 3,2008):1–14.
[2] 左腾.人脸识别技术综述[J].软件导刊,2017,16(2):182–185.
Zuo T.Face recognition technology review[J].Software Guide,2017,16(2):182–185.
[3] 张松兰.支持向量机的算法及应用综述[J].江苏理工学院学报,2016,22(2):14–17+21.
Zhang S L.A Survey of improved algorithm and application on support vector machine[J].Journal of Jiangsu University of Technology,2016,22(2):14–17+21.
[4] Samaria F S,Harter A C.Parameterisation of a stochastic model for human face identification[C]//Proceedings of 1994 IEEE workshop on applications of computer vision.IEEE,1994:138–142.
[5] Chang C C and Lin C j.LIBSVM - A Library for Support Vector Machines[EB/OL],(2019-04-04)[2019-12-17].https://www.csie.ntu.edu.tw/~cjlin/LIBSVM/.
[6] 周飞燕,金林鹏,董军.卷积神经网络研究综述[J].计算机学报,2017,40(6):1229–1251.
Zhou F Y,Jin L P,Dong J.Review of convolutional Neural Network[J].Chinese Journal of Computers,2017,40(6):1229–1251.
[7] Vladimir N.The Nature of Statistical Learning Theory[M].New York:Springer Press,2010.
[8] Zhao W,Chellappa R,Rosenfeld A,et al.Face Recognition:A Literature Survey[J].ACM Computing Surveys,2003,35(4):399–458.
[9] 卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016,31(1):1–17.
Lu H T,Zhang Q H.Applications of deep convolutional Neural Network in computer vision[J].Journal of Data Acquisition ＆ Processing,2016,31(1):1–17.
[10] 宋超峰,杨剑,宋文爱,等.改进的卷积神经网络架构[J].计算机工程与设计,2019,40(3):819–824.
Song C F,Yang J,Song W A,et al.Improved convolution neural network architecture[J].Computer Engineering and Design,2019,40(3):819–824.