The main goal of Human Action Recognition systems is to automatically identify and analyze the specific actions performed by human beings in videos. HAR is extensively used in many application areas, such as video storage and retrieval, robotics, intelligent video surveillance, and healthcare. The most challenging part of the action recognition problem is to process videos that contain actions. Other challenges associated with action recognition are cluttered backgrounds, illumination variations, occlusion, and intra- and inter-class similarities. Recently, Deep learning-based techniques have given the highest accuracy for recognizing human actions. In this paper, we discuss many deep learning-based techniques and pre-trained models proposed for recognizing human actions present in videos. Further, the performance comparison of various deep learning-based techniques with respect to obtained recognition accuracy is carried out. We also presented a brief overview of popular benchmark action datasets available for human action recognition. It is evident from the results produced by deep learning-based techniques that these methods are delivering excellent results for human action recognition and are capable of handling significant challenges associated with human action recognition.