Gender Classification with Human Gait Based on Skeleton Model

Abstract

Human gait walking skeleton model is proposed together with its implementation for gender classifications. The proposed model is based on morphological operations and is similar to the conventional skeleton model which allows calculations of joint angles of human body. Also gender classification method based on the proposed human gait walking skeleton model is proposed. Through experiments with the Class B dataset of Chinese Academy of Sciences (CASIA) silhouettes, it is confirmed that the proposed gender classification method utilizing human gait walking skeleton model allows discrimination between left and right legs even if a single camera acquired image is used. It is also confirmed that the proposed method allows estimation of joint angles accurately together with gender classification with high percent correct classification of 85.33% (it is 11.8% better classification accuracy comparing to the existing method).