Human action recognition based on skeleton features

Abstract

Based on human bone joints, skeleton information has clear and simple features and is not easily affected by appearance factors. In this paper, an improved feature of Gist, ExGist, is proposed to describe the skeleton information of human bone joints for human action recognition. The joint coordinates are extracted by using OpenPose and the thermodynamic diagram, and ExGist is used for feature extraction. The advantage of ExGist is that it can effectively characterize the local and global features of skeleton information while maintaining the original advantages of Gist feature. Compared with Gist, ExGist achieves better results on different classifiers. Additionally, compared with C3D and APTNet, our model also obtains better results with an accuracy rate of 89.2%.