Use this URL to cite or link to this record in EThOS:
Title: Combining sensors for activity recognition with uncertainty in Smart homes using Dempster-Shafer theory of evidence
Author: Liao, Jing
Awarding Body: University of Ulster
Current Institution: Ulster University
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
A population rapidly aging is a global problem. To meet the requirement of assistant living services for elderly person with a reasonable cost, one possible solution is through technology and the Smart Home, which is the main driving vision. Within the research community of Smart Homes, using multi-type sensors to monitor the resident daily activities has attracted more attention. This PhD project aims to build a set of evidential reasoning algorithms to integrate multi-sensor information for recognising human daily activities and to minimise the effect of uncertainty within data generated Smart Homes. In this study. two categories of uncertainty sources have been identified: the uncertainty derived from the envirorunent and the uncertainty from human activities. To deal with these kinds of uncertainty, two types of discounting rates have been suggested to accommodate for the uncertainty: hardware discounting and contextual discounting. To improve the accuracy of the Dempster-Shafer based reasoning method, a revised lattice-based structure was proposed and implemented in conjunction with incorporating the prior knowledge, such as historical activity patterns, into the reasoning process for activity recognition. Two weight factor algorithms, i.e. SWM and CWM, have been proposed to identify optima1 mapping strengths within the revised lattice structure. In this thesis, the relations between different layers, different activities and between sensors and activities were also investigated. To eva1uate the effectiveness of the proposed methods, they have been considered with three public datasets. The empirical results demonstrate that the Dempster-Shafer theory offers an easy way to represent two identified uncertainty information and recognize activity under uncertainty. The SWM algorithm is a better way for detecting the easy and nonnal human activities given that there is sufficient historical data. Otherwise, CWM is a good approach in instances of the lack of historical data and better for reasoning with complex activities.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available