Explainable pattern modelling and summarization in sensor equipped smart homes of elderly

In the next several decades, the proportion of the elderly population is expected to increase significantly. This has led to various efforts to help live them independently for longer periods of time. Smart homes equipped with sensors provide a potential solution by capturing various behavioral and physiological patterns of the residents. In this work, we develop techniques to model and detect changes in these patterns. The focus is on methods that are explainable in nature and allow for generating natural language descriptions. We propose a comprehensive change description framework that can detect unusual changes in the sensor parameters and describe the data leading to those changes in natural language. An approach that models and detects variations in physiological and behavioral routines of the elderly forms one part of the change description framework. The second part comes from a natural language generation system in which we identify important health-relevant features from the sensor parameters. Throughout this dissertation, we validate the developed techniques using both synthetic and real data obtained from the homes of the elderly living in sensor-equipped facilities. Using multiple real data retrospective case studies, we show that our methods are able to detect variations in the sensor data that are correlated with important health events in the elderly as recorded in their Electronic Health Records.

Name thesis:degree_name

Ph. D.

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Electrical and computer engineering (MU)

Grantor dc:publisher

University of Missouri--Columbia

Year dc:date.issued

2020

Author dc:creator

• Jain, Akshay

Advisor dc:contributor.advisor

• Keller, James

Statement dc:rights

• OpenAccess.

Language dc:language.iso

eng, English

DOI

https://doi.org/10.32469/10355/79503

OAI identifier oai:identifier

oai:mospace.umsystem.edu:10355/79503