Pedestrian characteristics are artificial definitions of meaning, which can be used as a soft biometric technology for visual inspection, and can also be used for pedestrian identification. The purpose of the gait function is to provide different images from low-level functions (such as HOG, LBP, Deep and SIFT) while changing the attributes of the target character. Due to changes in attributes, perspectives and viewing conditions, it can be regarded as high-level semantic information with more diverse meanings. Therefore, many computer vision programs integrate feature information into algorithms to improve performance. Although many studies have been conducted on the recognition of pedestrian characteristics due to factors such as angle changes, low illumination and low resolution, the recognition of pedestrian characteristics is still an unresolved problem. Traditional methods of identifying gait features usually focus on establishing strong features based on manual features, strong classifiers or feature associations. This includes HOG, SIFT, SVM or CRF types. The success of these traditional algorithms is far from meeting the requirements of practical applications.