Exploring the practicality of wearable gesture recognition

  • Jess McIntosh

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

The field of human-computer interaction has identified that gestures could be an important tool for natural interaction with computers. It has been envisioned that such an interaction modality could provide a convenient way of interacting with devices near to the person. In particular, hand gesture control could be useful for wearable devices, where interacting with these devices are difficult with other known modalities. Currently, smart watches and other wrist-worn devices are the dominant form factor for wearable devices, probably due to the socially acceptability of wrist-worn objects. Since the watch form factor is currently so dominant, it seems intuitive to embed the sensing technology for hand gesture recognition at this location.

This thesis analyses current wearable techniques for hand gesture recognition, paying particular attention to the practicalities of the techniques which are important for integration with a wrist-worn form factor device. Experimentation is conducted to improve existing techniques, attempting to address these practicality issues, with a focus on restricting the placement and size of the device to conform to the wrist. Further ergonomic and practical issues are uncovered through experimentation in EMG and ultrasonography, leading to a technical innovation to extend the capability of wearable infrared gesture sensing. Experience and knowledge gained from each chapter lets us advise that the only reasonably practical method that can be implemented at the moment is infrared. Finally, the analysis of methods also let us give insight on the most promising technologies (such as ultrasonography) and the main problems that hinder their practicality. One of the key problems that we find is that tightness of sensors against the skin are a big practicality concern, which is almost always a factor that is ignored in current research. Finally, we discuss additional robustness and cross-participant issues that remain challenging in all current techniques.
Date of Award23 Jan 2019
Original languageEnglish
Awarding Institution
  • The University of Bristol
SupervisorMike C Fraser (Supervisor)

Cite this

'