Face recognition using independent component analysis based on restricted mean field approximation
-
摘要: 针对有噪的ICA模型,提出一种有限制的平均场近似(restrictive m ean field approxim ation,RMFA)的算法来求解ICA模型参数和源信号的估计问题.在传统MFA-ICA算法的基础上,提出将ICA中的模型参数和源信号均限制为非负,目的是使得提取出的特征更独立,更利于识别.通过手写体数字和仿真模拟人脸图形以及ORL人脸数据进行实验,将RMFA-ICA算法与传统的ICA算法和无限制的MFA-ICA算法进行比较,对于手写体数字和仿真模拟人脸图形,RMFA-ICA算法能分离出更独立的特征,对于ORL人脸数据,其结果表明,利用RMFA-ICA算法明显优于传统ICA算法和无限制MFA-ICA算法识别结果.Abstract: Based on mean field approximation a new method is proposed to solve noisy ICA model,which can fairly well solve over-complete case,and estimate the independent source by restrict the non-negative mixing matrix and the non-negative sources.The experiments have been done for several different cases,such as digital images,simulated face graphics and ORL face database.The digital images and simulated face graphics experiments show the extraction features by the RMFA-ICA are more independent than that of using tradition ICA and unrestrictive MFA-ICA,the ORL face recognition experiments show the recognition ratio by the proposed method is greater than that of using the others methods.
-
-
[1] COMON P.Independent component analysis,a new concept[J].Signal Processing,1994,36(3):287～314.
[2] HYVARINEN A.Survey on independent component analysis[J].Neural Computing Surveys,1999,1(2):94～128.
[3] LEWICKI M S.Learning overcomplete representations[J].NeuralComputation,2000,12(2):337～365.
[4] PETERSON A.A mean field theory learning algorithm for neural networks[J].Complex Systems,1987,24(1):995～1019.
[5] ATTIAS H.Independent factor analysis[J].Neural Computation,1999,11(4):803～851.
[6] TANAKA T.A theory of mean field approximation[A].Advances in Neural Information Processing Systems[C].Massachustts:MIT Press,1999.351～357.
[7] HYVARINEN A,KARTHIKESH R.Sparse priors on the mixing matrix in ICA[A].Proc Int Workshop on ICA and BSS (ICA2000)[C].Helsinki:Helsinki University Press,2000.477～452.
[8] 刘直芳,游志胜,王运琼.基于PCA和ICA的人脸识别[J].激光技术,2004,28(1):78～81. [9] 刘直芳.人脸检测和识别的研究[D].成都:四川大学,2004.146～147.
计量
- 文章访问数: 3
- HTML全文浏览量: 0
- PDF下载量: 5
下载: