Real-time human action recognition using individual body part locations and local joints structure

作者：Liqiang Du, Hong Chen, Shuli Mei, Qing Wang · 年份：2016 · DOI：10.1145/3013971.3013974 · 被引用次数：4 · 研究领域：Human Pose and Action Recognition、Hand Gesture Recognition Systems、Gait Recognition and Analysis

In this paper, we present a novel approach for real-time human action recognition using local joints structure and body part locations. Individual body part locations are global features that ignore the local structure information of the human body joints, which is also essential for accurate action recognition. To cope with this problem, we propose local joints structure as a complement, and combine the two features for posture description in our method. We then perform classification using a combination of dynamic time warping, Fourier temporal pyramid representation and linear SVM. Experiments results on three action datasets show that the proposed representation outperforms many existing skeletal representations.