Uncorrelated discriminant simplex analysis for view-invariant gait signal computing

Human gait is a useful biometric signature and has recently gained growing interest from computer vision researchers. This interest is strongly driven by the need for automatic human identification and gender recognition at a distance in many surveillance applications. Existing human gait analysis methods, however, are sensitive to the view of the gait sequences, and their performances are poor when the view of the training gait sequences is different from that of the testing ones. In this paper, we propose a new supervised manifold learning algorithm, called uncorrelated discriminant simplex analysis (UDSA), for view-invariant gait signal computing. The aim of UDSA is to seek a mapping to project human gait sequences collected from different views into a low-dimensional feature subspace, such that intraclass geometrical structures are preserved and interclass distances of gait sequences are maximized simultaneously. Moreover, we impose an uncorrelated constraint to make the extracted features statistically uncorrelated. Experimental results are presented to demonstrate the efficacy of the proposed approach.