doi:10.1016/j.neucom.2005.11.004 
Copyright © 2006 Elsevier B.V. All rights reserved.
Letters
Palmprint recognition using FastICA algorithm and radial basis probabilistic neural network
Li Shanga, b, 
, De-Shuang Huangb, 
, , Ji-Xiang Dua, b, and Chun-Hou Zhenga, b,
aDepartment of Automation, University of Science and Technology of China, Hefei, Anhui 230026, China
bInstitute of Intelligent Machines, Chinese Academy of Sciences, P.O. Box 1130, Hefei, Anhui 230031, China
Received 6 September 2005;
revised 15 November 2005;
accepted 16 November 2005.
Communicated by R.W. Newcomb.
Available online 24 February 2006.
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Abstract
This paper proposes a novel and successful method for recognizing palmprint based on radial basis probabilistic neural network (RBPNN) proposed by us. The RBPNN is trained by the orthogonal least square (OLS) algorithm and its structure is optimized by the recursive OLS algorithm (ROLSA). The Hong Kong Polytechnic University (PolyU) palmprint database, which is pre-processed by a fast fixed-point algorithm for independent component analysis (FastICA), is exploited to test our approach. The experimental results show that the RBPNN achieves higher recognition rate and better classification efficiency than other usual classifiers.
Keywords: FastICA algorithm; Palmprint recognition; Radial basis probabilistic neural network (RBPNN); Recognition rate
Fig. 1. The independent basis image representation consisted of the coefficients, S, for the linear combination of independent basis images, U=(u1,u2,…,un), that comprised each palmprint image. FastICA representation is in S (S=(s1,s2,…,sn)).
Fig. 2. The factorial representation consisted of the independent coefficients, U, for the linear combination of basis images in A=(a1,a2,…,an), that comprised each palmprint image. FastICA factorial representation is in U (U=(u1,u2,…,un)).
Fig. 3. First 25 PC axes of the palmprint image set (columns of V), ordered left to right, top to bottom, by the magnitude of the corresponding eigenvalues.
Fig. 4. The first 25 basis images (rows of 
) that are obtained by FastICA architecture I. In this approach, the basis images are statistically independent.
Fig. 5. The first 25 basis images (columns of 
) obtained by architecture II. In this approach, the coefficients are statistically independent.
Fig. 6. The structure of radial basis probabilistic neural network.
Table 1.
Training and classification CPU time for the PolyU palmprint database
Abstract
Purchase PDF (344 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (6)
Purchase PDF (438 K)
