Abstract
Human motion analysis is crucial in fields such as healthcare, human-robot inter- action, and virtual reality. Traditionally, human movement is analysed based on human body or skeletal structure, which requires attaching sensors directly to the body when using a wearable motion capture device, a method that often limits user acceptance due to discomfort or inconvenience. With the advent of cutting-edge electronic textiles (e-textiles) technology, sensors can now be seamlessly integrated into clothing, enhancing user-friendliness. However, a challenge arises in understanding motion artefacts caused by the relative movement of clothing to the body.This thesis explores the performance of human motion recognition and prediction using sensorised garments. It presents comprehensive experiments employing statistical models based on the movement of loose clothing to predict body motion patterns, both in mechanically simulated environments and actual human behaviour.
These experiments cover a range of levels of similarity between the two movement patterns used for prediction tasks. In contrast to conventional approaches that rely on the human body or skeletal structure for analysis, our results suggest that using clothing motion for analysis can achieve significantly higher accuracy while requiring much less movement history to get satisfying motion recognition accuracy, particularly for tasks involving a high degree of similarity between the two movement
patterns.
To understand the reason behind this phenomenon, this thesis presents a probabilistic model that explains improved responsiveness and accuracy with fabric sensing from more discrimination information between movements recorded. Simulated and real human motion capture experiments with a number of participants confirm the model’s predictions, demonstrating that this phenomenon is accurately captured.
This thesis challenges the conventional belief that human motion analysis must rely on the human body or skeletal structure, demonstrating that garment motion may contain additional, useful information for human movement analysis. Moreover, not only might does this suggest the need for a revolutionary rethink of the design of motion capture technologies, it also has far-reaching potential implications in explaining the evolutionary and/or socio-cultural advantage of loose appendages (i) in nature, and (ii) traditional and modern costume.
| Date of Award | 1 Apr 2025 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Matthew Howard (Supervisor) |