Depth MHI Based Deep Learning Model for Human Action Recognition
Ye Gu, Xiaofeng Ye, Weihua Sheng · 2018
Human action analysis based on deep learning has become a hotspot in the fields of intelligent video. Recently, the approaches in depth-based human action recognition provide other ways to recognize human actions. The depth provide important supplementary information to improve the performance based on RGB stream. Meanwhile, the deep learning methods are effective for both RGB and depth features representation. In this paper, we use the deep learning model to learn the discriminative patterns for human action recognition from the depth-based motion history images (MHIs). This method is evaluated on both the 3D human action datasets RGBD-HuDaAct and NTU RGB+D. The experimental results show that our proposed approach achieves good accuracy for recognizing the indoor actions.