Abstract
In the past twenty years, there has been an increase in wearable technologies taking the form of everyday clothing, in both research and industry. This is due to the increasing accessibility of conductive textile materials, and the advancement of e-textiles fabrication techniques. A number of textile crafts have been exploited to enable embedded sensing and actuation in clothing and other textile objects. Large-scale weaving and knitting are the dominant approaches used, producing materials with sensing capability that take advantage of their inherent mechanical properties. Despite these advancements, there remain to be many challenges in designing and constructing wearable e-textile devices. They include sensor placement, networking of electronics, alongside considerations common in clothing design, including fit, comfort, and visual aesthetic. Through the work presented in this thesis, we demonstrate that solutions to these issues can be found by uniting garment construction practices with e-textiles fabrication.Grounded in standard fashion design and construction processes, we develop seam-based movement sensors which make use of the existing relationships between stitches, seams, and garment patterns. Through the work presented in this thesis, we identify the key design features of sensor seam performance. We begin with an exploration into fabricating seam-based movement sensors in bodices, iterating on our initial findings by extending our work to include sleeves. Through evaluations of both of these garment types, we show that sensor seam data can be used to differentiate between torso postures and arm gestures. The overall findings show that sensor seam architecture, seam placement, and stitch type are important design features that can be adapted to improve sensor seam sensitivity and robustness. Following this, we demonstrate the value in the application of sensor seams within the context of supporting patients during at-home physiotherapy practice. Using a sleeve form, we find that sensor seam data can be used to classify thirteen arm exercises with an average accuracy of 75.20%. We discuss how this value could be improved, and additional features necessary for a fully realised system. Finally, we suggest directions for future work based on our findings and in relation to existing gaps in research.
| Date of Award | 22 May 2024 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Michael Fraser (Supervisor) & Jason Alexander (Supervisor) |