Capture de mouvement par fusion de poses multivues pour une réadaptation en environnement virtuel

Authors: Bonenfant, Michaël
Advisor: Laurendeau, Denis
Abstract: Several works have shown the benefits of robotic assisted physical rehabilitation as well as rehabilitation in a virtual environment. Rehabilitation exercises are often performed using an expensive motion capture system based on markers attached to the patient. The use of this type of system requires time and expertise from a clinician. The project we are introducing aims to support such rehabilitation exercises in addition to avoid some of the problems of an expensive conventional motion capture systems. The developed system includes a cable mechanism integrated into a virtual environment providing haptic and visual feedback to the patient. The estimation of the human pose without marker is carried out in real-time to allow the analysis of the movement, the display of the patient avatar as well as its interaction with the virtual environment. The vision system exploit inexpensive 3D sensors to provide an affordable device for most rehabilitation clinics. That system allows enough reconfigurability to optimize the rehabilitation exercises to the needs of the patient. This thesis covers the stages of the system development, including the fusion approach based on multiple views based on particle filtering. The analysis of the results presented is promising, knowing that the system is also very flexible and adaptable to different contexts other than rehabilitation, such as the entertainment industry.
Document Type: Mémoire de maîtrise
Issue Date: 2017
Open Access Date: 24 April 2018
Permalink: http://hdl.handle.net/20.500.11794/27665
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
SizeFormat 
33227.pdf7.2 MBAdobe PDFView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.