Gait recognition based on structural gait energy image

Gait energy image (GEI) has been proven to be an effective gait recognition feature. It has good performance on most of public databases. However, GEI ignores most of gait motion information and doesn't contain enough human body structure information neither, which negatively affect its robustness in varying clothing and carrying conditions. We propose a new gait recognition method called structural gait energy image (SGEI), which combines the advantages of GEI and model-based methods. SGEI is generated by a fusion of foot energy image (FEI) and head energy image (HEI). A classifier fusion of GEI and SGEI is then conducted. Our method can cope with the clothing and carrying variations pretty well. We test it on CASIA(B) database and get a recognition rate of 89.29%, which is much higher than GEI whose recognition rate is 60.37% only. © 2013 Binary Information Press.