Recognition of quotidian activities in support of independent living using a single wrist-worn inertial measurement unit

Tools

Ortega Anderez, D. ORCID: 0000-0003-3571-847X, 2020. Recognition of quotidian activities in support of independent living using a single wrist-worn inertial measurement unit. PhD, Nottingham Trent University.

Preview

Text
Dario Ortega Anderez 2020.pdf - Published version
Download (4MB) | Preview

Abstract

The field of Ambient Assisted Living (AAL) is gaining increasing attention from the research community in recent years with the rapid present and future ageing of the population worldwide. This problem has been widely recognised as has the need for it to be addressed both from an economic and societal perspective. Assisted living environments incorporate technological solutions to create a better condition of life for older adults. However, in order to create a better condition of life, it is crucial to understand the specific needs of each individual. To this regard, self-assessment of daily activities has shown to be subjective and variable, presenting important discrepancies with those performed by clinicians.

The above challenges have fostered the search for alternative monitoring solutions, increasing the research efforts upon the field of Human Activity Recognition (HAR). A vast array of sensing devices, including ambient sensors, video cameras and wearable devices, has been employed for the automatic monitoring of a person in a home environment. However, the research focus is shifting towards wearable solutions, which avoid the privacy concerns related to the use of video cameras in a home environment while providing more intrinsic information about the user than ambient devices.

The focus of this research is the investigation of signal processing and machine learning techniques for the recognition of quotidian activities concerning self-neglect (a behavioural condition in which individuals, generally older people, disregard the attention, intentionally or un intentionally, of their basic needs). More precisely, the aimed group of activities include those concerning personal hygiene, namely handswashing and teeth brushing, as well as those directly related to dietary behaviour, namely eating and drinking.

The work undertaken in this thesis is divided into three different stages. First, given the continuous quasi-periodic behaviour of hands washing and teeth brushing, these are studied alongside a group of other quotidian activities which also exhibit continuity during their performance. These studies include the investigation of informative features for activity recognition as well as relevant classification models and signal processing techniques. In addition, a novel multi-level refinement approach is proposed as a way to improve the classification rate of those activities with lower inter-activity classification rate.

Second, a novel framework for fluid and food intake gesture recognition is developed. As opposed to the above activities, the nature of eating and drinking activities is neither static nor quasi-periodic. Instead, they are composed of sparsely occurring motions or gestures in continuous data streams. Given this characteristic, a novel signal segmentation technique, namely the Crossings-based Adaptive Segmentation Technique (CAST), is proposed to identify potential eating and drinking gestures while filtering out the remaining unwanted
segments of the signals. In addition, various feature descriptors, namely a Soft Dynamic Time Warping (DTW) gesture discrepancy measure and time series to image encoding techniques, as well as various deep learning architectures are explored to overcome the notable existing similarity between eating and drinking gestures.

The third stage of the work aims at the identification of meal periods through the analysis of the distribution of eating gestures along time using low-computational cost signal processing techniques, including a moving average and an entropy measure.

The novel computational solutions and the results presented in this thesis, demonstrate a significant contribution towards the recognition of quotidian activities in support of independent living.

Item Type:	Thesis
Creators:	Ortega Anderez, D.
Date:	July 2020
Divisions:	Schools > School of Science and Technology
Record created by:	Jeremy Silvester
Date Added:	12 Mar 2021 17:10
Last Modified:	31 May 2021 15:05
URI:	https://irep.ntu.ac.uk/id/eprint/42503