Identification of people at a distance using effective block list

Abstract

Gait energy image is an efficient gait descriptor for human gait recognition, but gait information in gait energy image is redundant and susceptible to shape scaling and drifting. To solve the problem, effective block list is proposed to express gait more effectively in this paper. For each row in the steady part of gait energy image, two blocks with max variation are selected to construct an effective block list. The same subject's difference sequence of effective block lists generally has a lower mean value, so it is used to measure the similarity between two subjects. Experimental results show that, the proposed effective block list is more efficient than gait energy image, and has a good ability to derive more effective feature extraction and expression methods. © Springer International Publishing 2013.