Int. J. Computational Science and Engineering, Vol. 19, No. 2, 2019 293
Copyright © 2019 Inderscience Enterprises Ltd.
Kernel-based tensor discriminant analysis with
fuzzy fusion for face recognition
Xiao-Zhang Liu* and Hang-Yu Ruan
College of Information Science and Technology,
Hainan University,
Haikou, China
Email: liuxiaozhang@gmail.com
Email: ruanhangyu258@163.com
*Corresponding author
Abstract: This paper proposes a novel kernel-based image subspace learning method for face
recognition, by encoding a face image as a tensor of second order (matrix). First, we propose a
kernel-based discriminant tensor criterion, called kernel bilinear fisher criterion (KBFC), which
is designed to simultaneously pursue two projection vectors to maximise the interclass scatter
and at the same time minimise the intraclass scatter in its corresponding subspace. Then, a score
level fusion method is presented to combine two separate projection results to achieve
classification tasks. Experimental results on the ORL and UMIST face databases show the
effectiveness of the proposed approach.
Keywords: kernel; tensor discriminant; bilinear discriminant; matrix representation; face
recognition.
Reference to this paper should be made as follows: Liu, X-Z. and Ruan, H-Y. (2019)
‘Kernel-based tensor discriminant analysis with fuzzy fusion for face recognition’, Int. J.
Computational Science and Engineering, Vol. 19, No. 2, pp.293–300.
Biographical notes: Xiao-Zhang Liu received his MS in Computational Mathematics and PhD
in Information and Computing Science both from Sun Yat-sen University, Guangzhou, China, in
2004 and 2010, respectively. He is currently a Professor from the College of Information Science
and Technology, Hannan University, Haikou, China. His research interests include pattern
recognition, image processing, and high dimensional data analysis.
Hang-Yu Ruan received his MS in Information and Computing Science from the Hubei Normal
University, Hubei, China, in 2016. He is currently pursuing his postgraduate from the College of
Information Science and Technology, Hannan University, Haikou, China. His research interests
include pattern recognition and image processing.
This paper is a revised and expanded version of a paper entitled ‘Kernel bilinear discriminant
analysis for face recognition’ presented at 2017 IEEE International Conference on Computational
Science and Engineering (CSE), Guangzhou, China, 21–24 July 2017.
1 Introduction
Nowadays, the development of data technology (DT) is
burgeoning (Kumar and Sahoo, 2017; Qiu et al., 2017; Zhou
et al., 2016), and face recognition is essentially a kind of
DT. During the past decades, subspace learning has been
one of the mainstream directions of the face recognition
field. Most conventional subspace learning techniques, such
as principal component analysis (PCA) Turk and Pentland
(1991) and linear discriminant analysis (LDA) Belhumeur
et al. (1997) are based on the so called vector-space model.
Under this model, the original two-dimensional (2D in
short) image data are reshaped into a one-dimensional (1D
in short) long vector by stacking either rows or columns of
the image. This vector-space model introduces the
following problems in practical applications. First, the
intrinsic 2D structure of image matrix is removed. As a
result, the spatial information stored in the 2D image is
discarded and not effectively utilised for representation and
recognition. Second, each image sample is modelled as a
point in a high-dimensional space, e.g., for an image of size
112 × 92, the commonly used image size in face
recognition, the dimension of the vector space is 10,304,
and the size of the scatter matrices is 10,304 × 10,304.
Obviously, a large number of training samples are needed to
get a reliable and robust estimation of data statistics. This
problem, known as curse of dimensionality, is often
confronted in real applications. Third, usually very limited
number of data are available in real applications such that
the small sample size (SSS) problem Fukunaga (1991)
comes forth frequently in practice.
To overcome the above drawbacks, efforts have been
made to seek to extract the 2D features directly from the