Date of Award

1-1-2019

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Electrical & Electronic Engineering

First Advisor

Prof. Dirk Pesch

Second Advisor

Dr. Piyush Agrawal

Abstract

Falls among the independently living elderly population are a major public health worry, leading to injuries, loss of confidence to live independently and even to death. Each year, one in three people aged 65 and older falls and one in five of them suffers fatal or non fatal injuries. Therefore, detecting a fall early and alerting caregivers can potentially save lives and increase the standard of living. Existing solutions, e.g. push-button, wearables, cameras, radar, pressure and vibration sensors, have limited public adoption either due to the requirement for wearing the device at all times or installing specialized and expensive infrastructure. In this thesis, a device-free, low cost indoor fall detection system using commodity WiFi devices is presented. The system uses physical layer Channel State Information (CSI) to detect falls. Commercial WiFi hardware is cheap and ubiquitous and CSI provides a wealth of information which helps in maintaining good fall detection accuracy even in challenging environments. The goals of the research in this thesis are the design, implementation and experimentation of a device-free fall detection system using CSI extracted from commercial WiFi devices. To achieve these objectives, the following contributions are made herein. A novel time domain human presence detection scheme is developed as a precursor to detecting falls. As the next contribution, a novel fall detection system is designed and developed. Finally, two main enhancements to the fall detection system are proposed to improve the resilience to changes in operating environment. Experiments were performed to validate system performance in diverse environments. It can be argued that through collection of real world CSI traces, understanding the behavior of CSI during human motion, the development of a signal processing tool-set to facilitate the recognition of falls and validation of the system using real world experiments significantly advances the state of the art by providing a more robust fall detection scheme.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Share

COinS