Simulation of a wireless sensor network for unobtrusively detecting falls in the home

Download files
Access & Terms of Use
open access
Embargoed until 2015-02-06
Copyright: Ariani, Arni
One serious issue related to falls among the elderly living at home or in a residential care facility is the ‘long lie’ scenario, which involves being unable to get up from the floor after a fall for 60 minutes or more. The first part of this thesis focuses on developing algorithms for unobtrusive falls detection using simulated responses from passive infrared (PIR) and pressure mat (PM) sensors, aimed at older subjects living alone at home. A Java-based wireless sensor network (WSN) simulator was developed. This simulation reads the room coordinates from a residential map, a path-finding algorithm (A*) simulates the subject’s movement through the residential environment. The fall detection algorithm was tested on 15 scenarios; three scenarios of ADL, and 12 different types of falls (four types of fall, each with three post-fall scenarios). A decision tree-based heuristic classification model is used to analyse the data and differentiate falls events from normal activities. The accuracy of the algorithm is 62.50%. The second part of this thesis focuses on addressing three remaining drawbacks of the previous algorithm and improving the robustness of the system. To solve the problem of the person continuing to move after falling, the potential effectiveness of using two PIR sensors at each location (which monitor the upper and lower halves of the room) is investigated. Graph theory concepts are used to infer how many people (or groups) are present in the environment, loosely track their movement/location, and monitor them independently for falls. This graph representation is also used to identify when someone leaves the residence. A revised fall detection algorithm, also based on a heuristic decision tree classifier model, is tested on 15 scenarios, each including one or more persons; three scenarios of ADL, and 12 different types of falls. The accuracy of the algorithm is 89.33%. Future work will focus on the investigation of the impact of using a more realistic (suboptimal) sensor characteristic on the performance of the designed fall detection algorithm, the fabrication of a hardware prototype and the preliminary implementation of this fall detection system in either a laboratory or real-world environment.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Ariani, Arni
Lovell, Nigel H.
Redmond, Stephen J.
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
Resource Type
Degree Type
PhD Doctorate
UNSW Faculty
download whole.pdf 1.88 MB Adobe Portable Document Format
Related dataset(s)