Human motion recognition has become one of the contemporary research areas of focus owing to the availability of sensors and accelerometers, cost-effective and less power consumption, live stream data, and improved computer vision, machine learning, artificial intelligence, and IoT.
   Human motion recognition helps identify and understand human actions, which are critical for many real-life applications. Human motion recognition has a broad range of applications, such as autonomous navigation systems, video surveillance, computer games, robotics, and human interactions.
   Modalities utilized for deep learning enabled human motion recognition: skeleton, depth, infrared sequence, point cloud, event stream, audio, acceleration, radar, and WiFi. RBM-based Models, Autoencoders, Convolutional Neural Networks, Recurrent Neural Networks, and Generative Adversarial Networks are the remarkable deep learning models employed for human motion recognition systems.
   Some of the most promising future research directions are multimodal and universal activity recognition, developing new deep models with less labeled data, crowdsourcing quality data for deep models, efficient deep learning models for resource-limited devices, hybrid networks, simultaneous exploitation of spatial-temporal-structural information and fusion of multiple modalities.
   Studies and surveys on human motion recognition using deep learning provide valuable information such as overview, preprocessing methods, modalities, challenges, future research scopes, benchmark datasets, and opportunities. Such survey and review papers are listed below;