Gait recognition by multidimensional representation for accumulated silhouette

Abstract

Gait is defined as a style of walking and the gait recognition is to recognize individual by using gait image sequence. Most studies about gait recognition use silhouettes which are extracted from gait image sequence because shape information included in the silhouette is more useful for recognition than others. In this paper, we propose gait recognition method using multidimensional representation for gait silhouettes. This paper focuses on the cyclic characteristics of gait. Thus we propose the method to form the accumulated silhouette regarding the cyclic characteristics and then describe those as multidimensional representation. In order to recognize individual using the multidimensional representation for the accumulated silhouette, we adopt tensor decomposition. We verify the superiority of the proposed approach via experiments with real gait sequences. © 2011 Springer-Verlag.