基于改进的ICA和RBF神经网络的人脸识别
|
摘要
为了提高人脸识算法的训练识别速度以及准确率,提出一种改进的人脸识别算法。将独立成分分析(independent component analysis,ICA)与径向基函数(radial basis function,RBF)神经网络相结合,利用ICA算法对人脸图像进行特征提取,采用牛顿迭代法提升其迭代性能;引入松弛因子,在保证收敛速度的前提下,放宽对初始权值选取的局限性。将特征信息作为RBF神经网络学习输入,采用监督聚类方法对神经网络进行构建和初始化,利用线性最小二乘法调整输出层连接权值,梯度下降法调整隐含层中心以及高斯宽带,通过训练学习获得最终的人脸识别分类结果。对比实验结果表明,改进的人脸识别算法训练速度和识别速度更快,准确率更高。
In order to improve the recognition speed and the accuracy of face recognition algorithm,an improved face recognition algorithm based on independent component analysis(ICA)and radial basis function(RBF)neural network is proposed in this paper.Firstly,ICA algorithm is used to extract features from face images,and the extended Newton iteration method is used to improve the iterative performance.The relaxation factor is introduced to relax the limitation of initial weight selection on the premise of ensuring the convergence speed.Then the feature information is used as the input for RBF neural network learning.The neural network is therefore constructed and initialized by using supervised clustering method.The linear least squares method is used to adjust the connection weights of the output layer,and the gradient descent method is for the hidden layer centre and Gaussian broadband.The final face recognition classification results is obtained through training.Experimental results show that the improved face recognition algorithm in this paper has faster training and higher recognition speeds,and therefore more accuracy on face recognition.
In order to improve the recognition speed and the accuracy of face recognition algorithm,an improved face recognition algorithm based on independent component analysis(ICA)and radial basis function(RBF)neural network is proposed in this paper.Firstly,ICA algorithm is used to extract features from face images,and the extended Newton iteration method is used to improve the iterative performance.The relaxation factor is introduced to relax the limitation of initial weight selection on the premise of ensuring the convergence speed.Then the feature information is used as the input for RBF neural network learning.The neural network is therefore constructed and initialized by using supervised clustering method.The linear least squares method is used to adjust the connection weights of the output layer,and the gradient descent method is for the hidden layer centre and Gaussian broadband.The final face recognition classification results is obtained through training.Experimental results show that the improved face recognition algorithm in this paper has faster training and higher recognition speeds,and therefore more accuracy on face recognition.
引文
[1]胡正平,陈俊岭.多层融合深度局部PCA子空间稀疏优化特征提取模型[J/OL].电子学报,2017,45(10):2383-2389[2017-03-13].http://kns.cnki.net/KCMS/detail/detail.aspx? dbcode=CJFQ&filename=RJDK201702057.DOI:10.3969/j.issn.0372-2112.2017.10.011.
[2]李康顺,李凯,张文生.一种基于改进BP神经网络的PCA人脸识别算法[J/OL].计算机应用与软件,2014,1(1):14-16[2017-03-14].http://kns.cnki.net/KCMS/detail/detail. aspx? filename=jyrj201401042&dbname=CJFD&dbcode=CJFQ.DOI:10.3969/j.issn.1000-386x.2014.01.041.
[3]李小薪,梁荣华.有遮挡人脸识别综述:从子空间回归到深度学习[J/OL].计算机学报,2017,9(2):634-639[2017-05-14].http://kns.cnki.net/KCMS/detail/detail.aspx?filename=JSJX201801011&dbcode=CJFQ.DOI:10.3788/lop53.111004.
[4] WANG Y.Face recognition based on ICA and SPSOELM[C/OL]//Information Technology, Networking,Electronic and Automation Control Conference.Chengdu:IEEE,2017(3)32:602-606[2017-05-14].https://doi.org/10.1109/ITNEC.2017.8284803.
[5] TEJA G P,RAVI S.Face recognition using subspaces techniques[C/OL]//Conference on Recent Trends In Information Technology.Chennai,Tamil Nadu,India:IEEE,2014:103-107[2017-05-16].http://doi.org/10.1109/ICRTIT.2012.6206780.
[6]杨信能,朱宝山.独立分量分析改进的图像去噪方法[J/OL].测绘科学,2015(3):159-165[2017-05-20].http://kns. cnki. net/KCMS/detail/detail. aspx?filename=CHKD201503028&dbcode=CJFQ.DOI:10.16251/j.cnki.1009-2307.2015.03.028.
[7]唐彪,金炜.多稀疏表示分类器决策融合的人脸识别[J/OL].电信科学,2018(4)26:33-37[2017-05-20].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFD&filename=DXKX201804004.DOI:10.11959/j.issn.1000-0801.2018010
[8]范群贞,刘金清.基于PCA/ICA的人脸特征提取新方法[J/OL].电子测量技术,2017,33(8):31-34[2017-05-21].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=DZCL201008008.
[9]黄璞,陈才扣.基于局部人脸图像的ICA人脸识别方法[J/OL].计算机工程与设计,2016,31(11):2550-2553[2017-05-21].http://kns.cnki.net/KCMS/detail/detail. aspx? dbcode=CJFQ&filename=SJSJ201011037.
[10]安大海,蒋砚军.基于BP神经网络的人脸识别系统[J/OL].软件,2015(12):375-379[2017-05-27].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=RJZZ201512019.DOI:10.16136/j.joel.2015.05.0588.
[11]HORIUCHI T,HADA T A.complementary study for the evaluation of face recognition technology[C/OL]//International Conference on Security Technology.Medellin:IEEE,2013(8):1-5[2017-05-14].http://doi.org/10.1109/CCST.2013.6922048.
[12]刘明珠,武琪,李昌.基于2DDPCA人脸识别算法[J/OL].电视技术,2016,40(1):122-126[2017-06-18].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=DSSS201601026.DOI:10.16280/j.videoe.2016.01.024.
[13]张金焕,吴进,宋骁,等.基于自适应相关滤波器的人眼定位改进方法[J/OL].电讯技术,2014(11):1538-1543[2017-06-18].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=DATE201411016.DOI:10.3969/j.issn.1001-893x/014.11.015.
[14]ZAFARUDDIN G M,FADEWAR H S.Face recognition:A holistic approach review[C/OL]//International Conference on Contemporary Computing and Informatics,Mysore,India:IEEE,2014:175-178[2017-06-22].http://doi.org/10.1109/IC3I.2014.7019610.
[15]SHEN M,CAO J T.Independent component analysis for face recognition based on two dimension symmetrical image matrix[J/OL].Control and Decision,2013:4145-4149[2017-05-14].http://doi.org/10.1109/CCDC.2012.6244664.
[16]李海朋,李晶皎,闫爱云.人脸识别中的遗传神经网络并行实现[J/OL].计算机科学,2015,S1(2):42-46[2017-06-22].http://kns.cnki.net/KCMS/detail/detail. aspx? dbcode=CJFQ&filename=JSJA2015S1040.DOI:10.16280/j.videoe.2016.01.008.
[2]李康顺,李凯,张文生.一种基于改进BP神经网络的PCA人脸识别算法[J/OL].计算机应用与软件,2014,1(1):14-16[2017-03-14].http://kns.cnki.net/KCMS/detail/detail. aspx? filename=jyrj201401042&dbname=CJFD&dbcode=CJFQ.DOI:10.3969/j.issn.1000-386x.2014.01.041.
[3]李小薪,梁荣华.有遮挡人脸识别综述:从子空间回归到深度学习[J/OL].计算机学报,2017,9(2):634-639[2017-05-14].http://kns.cnki.net/KCMS/detail/detail.aspx?filename=JSJX201801011&dbcode=CJFQ.DOI:10.3788/lop53.111004.
[4] WANG Y.Face recognition based on ICA and SPSOELM[C/OL]//Information Technology, Networking,Electronic and Automation Control Conference.Chengdu:IEEE,2017(3)32:602-606[2017-05-14].https://doi.org/10.1109/ITNEC.2017.8284803.
[5] TEJA G P,RAVI S.Face recognition using subspaces techniques[C/OL]//Conference on Recent Trends In Information Technology.Chennai,Tamil Nadu,India:IEEE,2014:103-107[2017-05-16].http://doi.org/10.1109/ICRTIT.2012.6206780.
[6]杨信能,朱宝山.独立分量分析改进的图像去噪方法[J/OL].测绘科学,2015(3):159-165[2017-05-20].http://kns. cnki. net/KCMS/detail/detail. aspx?filename=CHKD201503028&dbcode=CJFQ.DOI:10.16251/j.cnki.1009-2307.2015.03.028.
[7]唐彪,金炜.多稀疏表示分类器决策融合的人脸识别[J/OL].电信科学,2018(4)26:33-37[2017-05-20].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFD&filename=DXKX201804004.DOI:10.11959/j.issn.1000-0801.2018010
[8]范群贞,刘金清.基于PCA/ICA的人脸特征提取新方法[J/OL].电子测量技术,2017,33(8):31-34[2017-05-21].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=DZCL201008008.
[9]黄璞,陈才扣.基于局部人脸图像的ICA人脸识别方法[J/OL].计算机工程与设计,2016,31(11):2550-2553[2017-05-21].http://kns.cnki.net/KCMS/detail/detail. aspx? dbcode=CJFQ&filename=SJSJ201011037.
[10]安大海,蒋砚军.基于BP神经网络的人脸识别系统[J/OL].软件,2015(12):375-379[2017-05-27].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=RJZZ201512019.DOI:10.16136/j.joel.2015.05.0588.
[11]HORIUCHI T,HADA T A.complementary study for the evaluation of face recognition technology[C/OL]//International Conference on Security Technology.Medellin:IEEE,2013(8):1-5[2017-05-14].http://doi.org/10.1109/CCST.2013.6922048.
[12]刘明珠,武琪,李昌.基于2DDPCA人脸识别算法[J/OL].电视技术,2016,40(1):122-126[2017-06-18].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=DSSS201601026.DOI:10.16280/j.videoe.2016.01.024.
[13]张金焕,吴进,宋骁,等.基于自适应相关滤波器的人眼定位改进方法[J/OL].电讯技术,2014(11):1538-1543[2017-06-18].http://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&filename=DATE201411016.DOI:10.3969/j.issn.1001-893x/014.11.015.
[14]ZAFARUDDIN G M,FADEWAR H S.Face recognition:A holistic approach review[C/OL]//International Conference on Contemporary Computing and Informatics,Mysore,India:IEEE,2014:175-178[2017-06-22].http://doi.org/10.1109/IC3I.2014.7019610.
[15]SHEN M,CAO J T.Independent component analysis for face recognition based on two dimension symmetrical image matrix[J/OL].Control and Decision,2013:4145-4149[2017-05-14].http://doi.org/10.1109/CCDC.2012.6244664.
[16]李海朋,李晶皎,闫爱云.人脸识别中的遗传神经网络并行实现[J/OL].计算机科学,2015,S1(2):42-46[2017-06-22].http://kns.cnki.net/KCMS/detail/detail. aspx? dbcode=CJFQ&filename=JSJA2015S1040.DOI:10.16280/j.videoe.2016.01.008.