Deep Learning for Recognizing Human Activities Using Motion of Skeletal Joints

Abstract

 Together throughinventions “in consumer electronics, demands have increased for greater granularity in segregating and judging human daily activities. Additionally, the possibilities of machine learning, and particularly deep learning, is quitevisible as research excels in direction such as monitoring the elderly, and surveillance for detection of suspicious people and objects left abandoned. Even thoughmultiple implementations have been made to count Human Action Recognition (HAR) with body wear sensors, these provide worthless heft and burden onto the wearer for the same. Therefore, research predominantlyinsists upon image-based HAR system, bespoke frontier in the development of consumer products. This paper presents an intelligent human action recognition system capable of automatically analysing the human activities through depth sensors and mechanicsalong with human skeleton information, including together the concepts of image processing and deep learning” [1]. “Deep Learning is a subset collection of Machine Learning concerned where neural network algorithms inspired by human brain” [4] (what occurs spontaneously to human) learn from large amount of data through several layers for nonlinear transformation. The deep learning can process large number of features to increase the outcome accuracy.