Low Cost Intelligent Computer Vision based Assistive Technology for Elderly People

Dennis Cheung; Peter Joe; Muhammad Nadeem; Saleem Akhtar; M. Hassan Aslam

doi:10.22581/muet1982.2204.11

Author(s)

Dennis Cheung Department of Electrical, Computer, and Software Engineering, The University of Auckland,Computer, and Software Engineering,New Zealand

Peter Joe Department of Electrical, Computer, and Software Engineering, The University of Auckland,Computer, and Software Engineering,New Zealand

Muhammad Nadeem College of Engineering and Technology, American University of the Middle East,,Kuwait

Saleem Akhtar Department of Electrical and Computer Engineering, COMSATS University Islamabad, Lahore Campus,,Pakistan

M. Hassan Aslam Department of Electrical and Computer Engineering, COMSATS University Islamabad, Lahore Campus,,Pakistan

Abstract

The elderly population (aged above 65) has surpassed the number of children under five years and constitutes 10% of the population of the world, which is expected to reach up to 15% by 2050. A vast majority of the elderly suffer from some kind of chronic disease affecting their cognitive skills impairing their ability to spend a quality life independently. They have to rely on caretaker services for a quality life in a safer environment. However, caretaker services are quite expensive and it is hard for the health care organizations or the families to bear the expenses. 24/7 availability of caretakers is also an issue due to the shortage of workforce, failing to meet the ever-growing demand of the caretakers. This article presents a computer-vision based assistive system for the elderly to tackle this important problem. The system is capable of monitoring the environment as well as the activities of the subject in a standard room environment. It takes the video feed as an input and recognizes the activities of elderly people by using machine learning algorithms. The standalone system autonomously detects and generates real-time alerts for the caretaker, guardian or family member in case of any potential danger through an accompanying smartphone application. It also keeps records of the activities of the subject and generates reports which can be used by health professionals for further analysis and diagnosis. The proposed system was tested in a real-world environment and exhibited promising results for standing, sitting and resting, with a combined average accuracy of 83.41%. The proposed system is standalone, cost effective, and flexible for monitoring the activities of the elderly and is expected to enhance the safety and quality of life of the subject.

Publication Details

Page(s) 106-116

DOI 10.22581/muet1982.2204.11

Published Journal: Mehran University Research Journal of Engineering and Technology, Volume: 41, Issue: 4, Year: 2022

Request the full text from PASTIC

Keywords

Computer-Vision Human Detection Posture Recognition Machine Learning Neural Networks