Investigation on Human Activity Recognition using Deep Learning
S. Velliangiri, Iwin Thankumar Joseph, M Maravarman, P. Karthikeyan
Abstract
The goal of human activity recognition (HAR) is to describe a people action constructed on a set of sensor readings. Human activity recognition can be classified into two types economic and non-economic. Economic type is used to generate revenue. Non-economic type is used for mental satisfaction. Human activity recognition can be applied in the area of people work evaluations, elderly people care, convalescence, thief detections in a public place, intelligent homes and intelligent traffic. This paper discusses the deep learning model, merits, demerits and dataset used in human activity recognition. Finally, we have summarized essential challenges in HAR using deep supervised and deep unsupervised learning models.