Deep learning based human action recognition: A survey

Human action recognition has attracted much attentions because of its great potential applications. With the rapid development of computer performance and Internet, the methods of human action recognition based on deep learning become mainstream and develop at a breathless pace. This paper will give a novel reasonable taxonomy and a review of deep learning human action recognition methods based on color videos, skeleton sequences and depth maps. In addition, some datasets and effective tricks in action recognition deep learning methods will be introduced, and also the development trend is discussed.