Monitoring Indoor Living Spaces using Depth Information
Carl James Debono, Matthew Sacco, Joshua Ellul · 2020
Longer life expectancy is resulting in a steady increase in population that needs specific services to support their everyday routines. Public and private structures that provide services to these communities exist, however the increasing demands for service place these structures under stress and increased expenses. Assistive living systems can help reduce the demand and cost for these services by supporting the elderly at their homes, improving their quality of life in the process. In this paper we propose a solution that solely uses the depth information from RGB-D cameras to monitor the elderly within indoor living spaces. Deep learning on depth video data is used to detect the elderly and report the position to an application. This position information creates paths over time that can be monitored remotely by family members and caregivers to understand the behavior of the elderly and take appropriate action when needed. Experimental results show that the system manages to detect the person with an accuracy of 66.5% and a tracking accuracy of 59.1%.