Human gait recognition from videos is one of the promising research topics for analyzing human walking behavior. Spatio-temporal features and kinematics interesting points (three dimensional skeleton points) are the two key metrics in the gait examination. In general, input to gait recognition methods is categorized into 3 groups namely; two dimensional video-based, depth image-based and three dimensional (3D) skeleton-based methods. This work aims to present a survey on spatio-temporal and kinematic gait characteristics based on visual and 3D skeletal traits in RGB videos. A detailed insight on the various benchmarked gait databases, gait recognition representations based on model-based, model-free approaches and classifiers are presented in this review. Also, this paper investigates the performance metrics, application areas and covariate factors that influence the gait recognition process. Finally, the paper outlines the future perspective of gait recognition system based on kinematic joint points.

从视频中识别人类步态是分析人类步行行为的有前途的研究课题之一。时空特征和运动学兴趣点（三维骨架点）是步态检查中的两个关键指标。一般来说，步态识别方法的输入分为三类：基于二维视频、基于深度图像和基于三维 (3D) 骨架的方法。这项工作旨在对基于 RGB 视频中的视觉和 3D 骨骼特征的时空和运动步态特征进行调查。本综述详细介绍了各种基准步态数据库、基于模型、无模型方法和分类器的步态识别表示。此外，本文研究了性能指标，应用领域和影响步态识别过程的协变量因素。最后，论文概述了基于运动学关节点的步态识别系统的未来发展前景。