ORIGINAL ARTICLE
Human face recognition based on ensemble of polyharmonic
extreme learning machine
Jianwei Zhao
•
Zhenghua Zhou
•
Feilong Cao
Received: 21 February 2012 / Accepted: 29 January 2013
Ó Springer-Verlag London 2013
Abstract This paper proposes a classifier named
ensemble of polyharmonic extreme learning machine,
whose part weights are randomly assigned, and it is har-
monic between the feedforward neural network and poly-
nomial. The proposed classifier provides a method for
human face recognition integrating fast discrete curvelet
transform (FDCT) with 2-dimension principal component
analysis (2DPCA). FDCT is taken to be a feature extractor
to obtain facial features, and then these features are
dimensionality reduced by 2DPCA to decrease the com-
putational complexity before they are input to the classifier.
Comparison experiments of the proposed method with
some other state-of-the-art approaches for human face
recognition have been carried out on five well-known face
databases, and the experimental results show that the pro-
posed method can achieve higher recognition rate.
Keywords Human face recognition Ensemble of
polyharmonic extreme learning machine Fast discrete
curvelet transform 2-Dimension principal component
analysis
1 Introduction
Human face recognition, as one of the most important
methods for the identification, has been widely studied
recently. Compared with some other approaches in bio-
metrics, such as fingerprint, iris, and DNA recognition,
human face recognition may not have a better level of
accuracy, but it is inexpensive, convenient, and hassle-free.
Therefore, human face recognition has widespread appli-
cations, such as smart cards, telecommunication, database
security, medical records, digital libraries, and so on [1].
Human face recognition system includes two key steps:
feature extraction and classification. Feature extraction is
to give an effective representation of the face images to
decrease the computational complexity of the classifier,
which can greatly enhance the performance of the human
face recognition system. While classification is to distin-
guish those features with a good classifier. Therefore, in
order to improve the recognition rate of a human face
recognition system, it is crucial to find a good feature
extractor and an effective classifier.
Recently, many methods for dimensionality reduction
have been proposed to improve the accuracy and speed of a
human face recognition system [2–5]. Kirby et al. [6] and
Turk et al. [7] employed principal component analysis
(PCA) to obtain a lower dimensional representation of the
human face image. As one of classical methods for
dimensionality reduction, PCA provides effective approx-
imation, but it suffers from high combinational load and
poor discriminatory power [8]. To eliminate these limita-
tions of PCA, researchers have proposed some other
techniques for dimensionality reduction, such as indepen-
dent component analysis (ICA) [9], linear discriminant
analysis (LDA) [10], Kernel principal component analysis
(KPCA) [11], Kernel linear discriminant analysis (KLDA)
[12], and so on. However, converting matrices into vectors
often results in a high-dimensional vector space, in which it
is difficult to evaluate the covariance matrix accurately due
to its large size and the relatively small number of training
samples. Furthermore, it is time-consuming to compute the
eigenvectors of a large size covariance matrix. In order to
J. Zhao Z. Zhou F. Cao (&)
Department of Mathematics, China Jiliang University,
Hangzhou 310018, Zhejiang Province,
People’s Republic of China
e-mail: feilongcao@gmail.com
123
Neural Comput & Applic
DOI 10.1007/s00521-013-1356-4