Different Techniques for Human Activity Recognition

R. Raj, A. Kos (AGH Univ. of Science and Techn., Poland)

Human activity recognition (HAR) in live videos has become an essential and important research topic in computer vision. RHA is widely used in different fields such as healthcare, robot learning, intelligent surveillance system, human-computer interactions (HCI), and many more. Recognition of activities in live or normal videos includes a huge amount of data are required to be processed that is why it is a tough task. In recent decades, researchers have developed various models using artificial intelligence and especially deep learning (DL) with multiple input sensor paradigms. This paper provides a comprehensive review of recent models of deep learning for HAR based on input sensor paradigm, dataset, feature extraction from the dataset, preprocessing of data, classification of input data, and accuracy.