Monitoring Elderly Falls in the Home Using Computer Vision

Smith, Jabin (2021) Monitoring Elderly Falls in the Home Using Computer Vision. [USQ Project]

Text (Project)
SMITH Jabin dissertation_redacted.pdf

Download (12MB) | Preview


The population of the world is getting older on average and there are some suggestions that by 2066, the population of Australians over the age of 65 could be as high as 20%. Researchers have also found that more and more elderly people are choosing to spend their twilight years living in their own homes, a phenomenon dubbed ‘aging in place’. One of the biggest hospitalisation injuries for the elderly is that which is caused by falling. Falls account for 40% of injury related death in the elderly and researchers have also found that the likelihood of death is vastly increased if the elderly person is unable to get back up after they fall. A term called the ‘long lie’ is used to explain this phenomenon and the aim of this research will be to create a fall detection technique that can detect the occurrence of the long lie and thus, prevent it.

This project has explored different types of research already conducted on capturing falls within the home and analysed what types of technology were used. Of these fall detection techniques, this project is focussed on using computer vision to detect falls. The method of image processing used for this project was foreground extraction and the fall detection algorithm utilised shape analysis of the foreground mask to determine if a fall occurred. This project explored the reliability and effectiveness of this type of algorithm and, after several iterations, an algorithm is presented that was able to detect the occurrence of a fall in most scenarios in the datasets provided. The final algorithm used a foreground detector provided by MATLAB’s computer vision toolbox in conjunction with a blob detector that was able to analyse the foreground mask and produce outputs based on the mask. The fall detection algorithm then uses a combination of these outputs to determine if a fall has occurred. This algorithm is unique in that it uses a state-based system where if any of the fall conditions exist, the state will change to a fall state. The system will remain in this state until the algorithm has detected that the person has returned to the upright position. This provided a means to ensure that the algorithm could detect the occurrence of a long lie. There were, however, many occasions when the algorithm either incorrectly identified a fall or did not detect the fall at all. This was mainly due to problems with the foreground extraction which led to the conclusion that if a suitable foreground extraction is not produced, then fall detection is not going to be reliable.

This research also explores the influence of ethics and what impact computer vision systems have on society. It will find that the major concerns surrounding computer vision systems and other artificial intelligence (AI) is the threat to security and privacy. It was also discovered that very little has been done to produce ethical standards in the AI industry and that it is incumbent on all stakeholders to ensure that a moral set of guidelines are produced for the entire lifecycle of an AI product. The research will also show that trust in AI is achieved by ensuring that users understand, not only the benefits of the technology, but also the risks associated with it and how these risks are mitigated.

Statistics for USQ ePrint 51841
Statistics for this ePrint Item
Item Type: USQ Project
Item Status: Live Archive
Faculty/School / Institute/Centre: Historic - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering (1 Jul 2013 - 31 Dec 2021)
Supervisors: Low, Tobias
Qualification: Bachelor of Engineering (Honours) (Instrumentation, Control and Automation)
Date Deposited: 03 Jan 2023 04:43
Last Modified: 26 Jun 2023 02:10
Uncontrolled Keywords: ageing, fall, computer, artificial intelligence, elderly, computer vision, algorithm, fall detection, ethics

Actions (login required)

View Item Archive Repository Staff Only