Depth Estimated History Image based Appearance Representation for Human Action Recognition
Suraj Prakash Sahoo, Samit Ari · 2019
Depth of a human action scene is important to distinguish human action from its background. Therefore, it helps to describe the appearance of the action. In this work, depth history image (DHI) is proposed by estimating depth from action frames. The DHIs are then applied to AlexNet to finetune the weights of the pre-trained deep learning architecture. To recognize the closely related actions, DHI alone is not sufficient. The 3D projected planes are extracted and trained separately on AlexNet for this purpose. Two types of projected planes are extracted in this work such as XT plane or side view and YT plane or top view of the action videos. The scores from both the learning techniques are fused to provide the final recognition score. The proposed HAR technique is evaluated on well established KTH and Weizmann human action datasets and the results suggest that the proposed HAR is better compared to most of the state-of-the-art methods.