Simple and efficient pose-based gait recognition method for challenging environments

Abstract

Listen

Translate article

Gait is a biometry characterized by the identification of individuals by the way they walk. It is recently gaining evidence because it can be collected at distance and does not require subject cooperation, which is desirable on surveillance scenarios. Despite these advantages, the literature reports challenging situations where gait recognition is not accurate and although exist works that try to address these problems, most of them uses silhoettes, which carry appearance information that confounds with gait. Because of this limitation, a pose estimation method that use information of frames is employed for gait recognition and a multilayer perception, called PoseFrame, is created. As the focus of gait is the classification of a whole walking sequence, the results based on the frames are temporally aggregated for final classification. The method is tested on CASIA Dataset A, having accuracy above other pose-based works; and on CASIA Dataset B, achieving the best results in some situations. An ablation study is also performed, finding that the arms and feet are the most important body parts for gait recognition.