2016 IEEE International Conference of Online Analysis and Computing Science (ICOACS 2016)

70 / 2015-11-19 13:58:10

Orthogonal bidirectional discriminant supervised locality preserving projection for face recognition

face recognition;discriminant supervised information;orthogonal bidirectional projection

Two-dimensional discriminant locality preserving projection (2DDLPP) and two-dimensional discriminant supervised LPP (2DDSLPP) are two effective two-dimensional projection methods for dimensionality reduction and feature extraction
of face image matrices. Since 2DDLPP and 2DDSLPP preserve the local structure information of the original data and exploit the
discriminant information, they usually obtain good recognition performance. However, 2DDLPP and 2DDSLPP only employ
single-sided projection, and thus the generated low dimensional data matrices have still many features. So the bidirectional
discriminant supervised locality preserving projection (BDSLPP) was proposed to overcome this problem. In this paper, by
combining the discriminant supervised LPP with the orthogonal bidirectional projection, we propose the orthogonal bidirectional discriminant supervised LPP (OBDSLPP), in which the left and right projection matrices for OBDSLPP are orthogonal. In this algorithm, to get the matrices, one trace ratio optimization problem are required to be solved. Experimental results show
that the proposed OBDSLPP achieve more higher recognition accuracy than 2DDLPP,2DDSLPP and BDLPP,respectively.