UWSpace is currently experiencing technical difficulties resulting from its recent migration to a new version of its software. These technical issues are not affecting the submission and browse features of the site. UWaterloo community members may continue submitting items to UWSpace. We apologize for the inconvenience, and are actively working to resolve these technical issues.
 

WatchTrace: Design and Evaluation of an At-Your-Side Gesture Paradigm

Thumbnail Image

Files

Siddhpuria_Shaishav.pdf (18.77 MB)

Date

2017-11-16

Authors

Siddhpuria, Shaishav

Journal Title

Journal ISSN

Volume Title

Publisher

University of Waterloo

Abstract

In this thesis, we present the exploration and evaluation of a gesture interaction paradigm performed with arms at rest at the side of one's body. This gesture stance is informed persisting challenges in mid-air arm gesture interactions in relation to fatigue and social acceptability. The proposed arms-down posture reduces physical effort by minimizing the shoulder torque placed on the user. While this interaction posture has been previously explored, the gesture vocabulary in previous research has been small and limited. The design of this gesture interaction is motivated by the ability to provide rich and expressive input; the user performs gestures by moving the whole arm at the side of the body to create two-dimensional visual traces, as in hand-drawing in a bounded plane parallel to the ground. Within this space, we present the results of two studies that investigate the use of side-gesture input for interaction. First, we explore the users' mental model for using this interaction by conducting an elicitation study on a set of everyday tasks one would perform on a large display in public to semi-public contexts. The takeaway from this study presents the need for a dynamic and expressive set of gesture vocabulary including ideographic and alphanumeric gesture constructs that can be combined or chained together. We then explore the feasibility of designing such a gesture recognition system using commodity hardware and recognition techniques, dubbed WatchTrace, which supports alphanumeric gestures of up to length three, providing a vibrant, dynamic, and feasible gestural vocabulary. Finally, we explore potential approaches to improve the recognition through the use of adaptive thresholds, n-best lists, and changing reject rates among other conventional techniques in the field of gesture classification.

Description

Keywords

gestures, wearables, smartwatch

LC Keywords

Citation

URI

http://hdl.handle.net/10012/12627

Collections

Theses
Computer Science