University of Victoria (Canada)
Computer vision-based analysis of human daily actions using Hidden Markovian Models
Abstract
dc:description.abstractThe study of human motion from a medical standpoint has traditionally involved the use of marker-based motion tracking systems, as well as other sensory devices. This equipment is often expensive, has low-portability, and might even influence tracking results by distracting or otherwise inhibiting a subject's normal motion performance. In comparison, non-marker-based tracking methods are less costly, easier to move and set up, and requires no markers or other devices. However, previous work in silhouette-based human motion analysis is typically focused on the classification of activities or the identification of subjects, neither of which are much use to medical professionals. We propose a merging of these two research fields. By applying silhouette-based motion tracking to the problem of motion performance analysis, we have developed a new method which can reliably and accurately model human motions and detect abnormalities. Our approach, which is based on Hidden Markovian Models with continuous observation probabilities, creates standardized models to represent common human motions. These models are then used as a basis for further analysis. We have extensively tested the proposed method with a custom designed database that takes into account speed related and subject related variations of motion performance.
Degree
thesis:*- Level thesis:degree_level
- Master of Applied Science M.A.Sc.
- Department thesis:degree_department
- Department of Electrical and Computer Engineering
- Year dc:date.issued
- 2011
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Beugeling, Trevor Robert John
- Advisor dc:contributor.supervisor
-
- Branzan-Albu, Alexandra
Subjects
dc:subject × 2Rights
dc:rights- Statement dc:rights
-
- Available to the World Wide Web
- Language dc:language.iso
- en, English
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- http://hdl.handle.net/1828/3293