Graduation Year

2020

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Industrial and Management Systems Engineering

Major Professor

Ali Yalcin, Ph.D.

Committee Member

Carla VandeWeerd, Ph.D.

Committee Member

Mingyang Li, Ph.D.

Committee Member

Yu Zhang, Ph.D.

Committee Member

Lu Lu, Ph.D.

Keywords

Age in Place, Health Monitoring, Mobility, Predictability, Smart Homes

Abstract

The world’s population is rapidly aging and the increasing demand for home and health care services from this aging population brings unprecedented challenges to the economy and society. Ambient-assisted smart homes, residences equipped with ambient sensors to monitor the resident’s daily activities in a continuous and unobtrusive way, present great potential to manage the growing care service needs of this older population segment, and enable them to age-in-place.

Despite growing research, using ambient sensor data from private homes to monitor daily activities, health and wellness still faces significant challenges. To study ambient sensor data from private homes where annotated data is unavailable and sensor layouts are variable, we proposed a novel two-phase location and status estimation algorithm to monitor health and wellness related metrics from ambient sensor data. The proposed algorithm is highly accurate as validated by a mobile app that prompts participants with questions about the estimated time of their daily activities. The outputs of this algorithm facilitate the visualization and examination of older adults’ daily patterns and activities, and through case studies, we show that it has the potential to be used with a wide range of ambient sensor networks with any mix of motion sensor types.

We also studied human mobility in private homes. Understanding human mobility is fundamental and critical for the design of context-aware assistive services in smart homes. We represent the resident’s movement trajectory based on ambient motion sensor data and use the entropy rate to quantify the regularity of the resident’s mobility patterns to estimate an upper bound of predictability. A change point detection algorithm based on penalized contrast function is used to identify the time periods when the data does not completely reflect the resident’s activities due to the presence of visitors and sensors system faults. Experimental results using data collected from 10 private homes over periods of 178 to 713 days show that human mobility at home is not completely random but regular and highly predictable independent of variations in floor plans and individual daily routines, which is consistent with the conclusions about human mobility in outdoor environments.

Finally, we summarize and analyze records in maintenance logs and bi-weekly assessments about changes and disruptions in ambient sensor data collected from private homes, and suggest potential research directions for the design of stable and reliable health and wellness monitoring systems using ambient sensor systems.

Download

Share

COinS