Delva, Mona Lisa - Hand Gesture Identification in Older Adults using Force-Myography...

This thesis has been approved for inclusion in the SFU Library.
Publication of this thesis has been postponed at the author's request until 2018-06-20.
Term: 
Summer 2017
Degree: 
M.A.Sc.
Degree type: 
Thesis
Department: 
School of Engineering Science
Faculty: 
Applied Sciences
Senior supervisor: 
Carlo Menon
Publishing Documentation
Postponement release date: 
Wed, 2018-06-20
Thesis title: 
Hand Gesture Identification in Older Adults using Force-Myography
Given Names: 
Mona Lisa
Surname: 
Delva
Abstract: 
The projected increase in the proportion of seniors in society has prompted the growth of senior-technologies that support aging-in-place. The aim of this thesis to explore the suitability of Force Myography (FMG) for hand gesture identification in aging populations to complement other technologies that promote aging-in-place and to investigate the practical considerations for implementation. Characteristics of using FMG with seniors (aged 60+ years old) was first determined with a protocol involving five seniors and five non-seniors. Participants were invited to don a custom FMG device and perform a series of stationary hand gestures while being guided by a virtual user interface. The interface provided online image instructions of the required gesture, as well as visual feedback of successful gesture identification. Participants also performed household activities based tasks in a self-selected manner. On average, seniors completed specified hand gestures within 1.4 seconds of online instruction, with inadvertent identification of control gestures during household tasks lasting at most 1.45 seconds. Although these times were comparable non-senior participants, seniors demonstrated increased variability. Lastly, online accuracies for gesture classification only reached 75% compared to the 91% of non-senior participants. Considering the results of the first study, a follow up study was performed with a larger recruitment pool focusing on intrinsic user features that influence the variability in FMG acquisition and modelling. The results demonstrate that age and gender associated differences in band tightness, grip strength and ratio of skinfold thickness to forearm circumference account for at most 30% of the variability in FMG responsiveness, translating to 7% to 30% of the variability of model test accuracy. Intrinsic user features also influenced the severity that functional noise (the affect of unintended movements) had on classification. Results also revealed that variables independent of the user, such as band removal, contribute significantly to declines in testing accuracy, where declines ranged from 28% to 96%. Finally, results also showed that methods of FMG modelling typically encountered in the literature shows limited effectiveness during non-static activity.
Keywords: 
activities of daily living; aging; rehabilitative and assistive technology; human factors; sensors/sensor application; force myography
Total pages: 
166