**2. State-of-the-Art**

Currently, the recognition of human activities has become more popular in the computer vision committee, and this interest is translated by many applications into more activities such as surveillance, video games, physical rehabilitation, etc. In this case, we can distinguish three emerging branches in the research on the recognition of activities: (1) depth-based representation, (2) skeleton-based representation, and (3) RGB-D-based development. In this section, we will go back to the existing work of recognizing human activities captured by depth cameras and describe in more detail the literature on the specific shared structures of learning for the recognition of activities.
