Year: 2015
Type: dissertation
Abstract: This dissertation proposes smart home-based intelligent techniques that perform automated assessment of a resident's well-being by monitoring their behavior inside the home. We hypothesize that the everyday behavior of smart home residents can be estimated by tracking residents' activities using smart home sensors and that machine learning algorithms can predict their cognitive and physical health utilizing behavioral information.\nWe first describe a cross-sectional study where we compare behavior differences across an entire population sample to assess activity quality and the individual's cognitive health. For this study, we introduce a machine learning-based framework for assessing the quality of eight different activities and one complex activity with interweaved sub-activities, called the Day Out Task. We compare our automated assessment of task quality with direct observation scores and assess the ability of machine learning techniques to classify an individual's cognitive health using the same machine learning-based framework.\nWe then describe a longitudinal study where we use an individual as their own baseline to identify behavioral changes that can predict cognitive health and mobility. We first introduce a Clinical Assessment using Activity Behavior (CAAB) approach to model a smart home resident's daily behavior and predict the corresponding standard clinical assessment scores utilizing longitudinal smart home sensor data. CAAB extracts statistical features that describe characteristics of a resident's daily activity performance and trains the machine learning algorithms to predict the standard clinical assessment scores. We then introduce an activity curve to represent an abstraction of an individual's normal daily routine based on automatically recognized activities. We develop algorithms to detect changes in behavioral routines by comparing activity curves and use these changes to analyze the possibility of changes in cognitive or physical health. We evaluate all of our algorithms using real-world longitudinal smart home sensor data. We conclude that it is possible to assess the health and well-being of a smart home resident utilizing smart home sensor data and machine learning algorithms.
Author: Prafulla N. Dawadi
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2532671952