Human action recognition using meta learning for RGB and depth information

In this paper, we propose an efficient human action recognition technique, which utilizes Depth and RGB information of the scene. Our proposed technique, first builds a pair of classifiers based on RGB and depth information to independently predict the actions within a scene. Then, the obtained results from these classifiers are combined to achieve high accuracies in human action recognition. Our experimental results show that using an efficient amalgamation of depth-based and RGB-based classifiers improves human action recognition in smart home applications.