Abstract
Gait analysis traditionally occurs within laboratory settings, failing to identify real-life gait complexities and only offering short-term observations. This thesis aims to investigate the use of smart insoles as a cost-effective and non-invasive alternative solution for gait analysis in daily life, supporting clinicians in the assessment of gait disorders by providing evidence-based insights into the behaviour and patterns of an individual’s gait. A comprehensive analysis was conducted to determine the optimal smart insole settings and sensors for accurate activity recognition, highlighting the integration of pressure and inertia sensors as the optimal configuration, which led to the development of a novel Deep learning algorithm for real-time activity classification. Such an algorithm allows the system to determine the action carried out by an individual and apply the best strategy to evaluate their gait patterns. By exploiting the information provided by smart insoles and extracting the related statistical features, a study for the early diagnosis of Parkinson’s Disease was conducted by comparing different machine learning models, demonstrating high accuracy and precision. Although this solution provided a valid overview of a subject’s status, clinicians need quantitative information to develop tailored rehabilitation plans. Therefore, a novel algorithm was designed for the identification of gait phases, presenting an unsupervised finite state machine, which allowed the extraction of gait kinematic and kinetic parameters. Finally, to evaluate the algorithms and system developed, a feasibility analysis was carried out by applying smart insoles in neurological and musculoskeletal disease assessment during exoskeleton-assisted rehabilitation and in evaluating gait symmetry in lower-limb amputees. This thesis lays the foundation for the integration of AI-enhanced smart insoles for the assessment of gait disorders in daily life. The case studies highlight how these systems provide quantitative information on gait, fostering a comprehensive assessment for clinicians, and improving the provision of personalized healthcare.Thesis is embargoed until 30th June 2026
| Date of Award | Jun 2024 |
|---|---|
| Original language | English |
| Supervisor | Huiru (Jane) Zheng (Supervisor) & Haiying Wang (Supervisor) |
Keywords
- smart insoles
- artificial intelligence
- human activity recognition
- gait analysis