This work is funded by EUR Digisport and will enable to validate strong scientific knowledge on computer vision and robotic on a societal issue. It allies theoritical and practical skills to be evaluated by simulation and experimentations.
The thesis starts in October 2022
Wheelchair sports such as soccer, basketball are becoming increasingly popular for athletes with disabilities. In France, the perspective of the 2024 Paralympic Games highlights these sporting disciplines that nowadays appear much more accessible. Sport plays an important role in the well-being of individuals: it is known to improve the self-esteem and the autonomy of people with disabilities, as well as the social inclusion . Practising sport has also a positive impact on rehabilitation and motivation.
The proposed PhD aims then to offer technological tools to improve performances of users of power wheelchairs: analysis of sessions, training aid, rehabilitation tools through sport. To this aim, we propose to design solutions compatible with the virtual reality power wheelchair simulator developed at INSA Rennes as part of the European Interreg ADAPT project. Simulators have already been proven to be efficient in the context of rehabilitation   and we want to extend the study of the possibilities offered by these platforms to sport situations. Concretely, our project seeks to play virtual or acquired soccer game scenarios on the simulator for training and rehabilitation sessions. Thus, the idea is to be able to replay game sequences for the analysis and improvement of individual performances. To do this, it is necessary to detect the static and dynamic elements of the environment and to locate them relatively to the wheelchair. Several kinds of system acquisition will be considered such as motion capture systems in order to provide motion models for virtual scenarios. To get information from all players during a match in a collective game, it is necessary to also consider vision to detect players, ball and equipment. It will be also required to add vision sensors on the wheelchair when possible to get the point of view of the player for feedback. The robustness of the solution with respect to the vibrations of the chair in action should be evaluated.
Specific demands have been raised by soccer trainees for people using a wheelchair of the team of Rennes (http://www.handisport-rennes-club.org/) and we will focus on those use cases. They asked for some feedback tools so they can study afterwards the vision field of each player depending their disabilities during a match. This a augmented reality application that requires wheelchairs tracking, identification and localisation in 3D. They also need video analysis to improve player placement during specific game phases. As far as virtual scenarios are concerned, they need training tools to help players to control the ball for example. As a specific powered wheelchair is used, it is tiring, discouraging in the worst case, to play efficiently when beginning. Let note such powered wheelchairs as the Strike force model are a lot faster that the ones used in everyday life and have a different design.
The augmented reality application to display the field of view of each player requires to fuse occluded body pose estimation and precise wheelchair tracking and localization. It will be the core of the PhD thesis.
This work requires to tackle the issue of modeling power wheelchair behaviour during a game using the work previously done in the Interreg ADAPT project. The motion capture system will be used to evaluate the validity of the approach.
The candidate will improve the state-of-the-art detectors. It requires building a dedicated database (images with localisation of the wheelchair) with players on wheelchairs. The motion capture system will be used to provide the ground-truth. Let note that such a database could be made public to ease action recognition of disabled people with wheelchair in other contexts.
The obtained detectors will then help to provide a tracker for all players in a game sequence. The output of such a tracker will be used to provide 3D data for an immersive replay using the local IMMERSIA platform (https://www.irisa.fr/immersia/services/).
As presented, a local team of soccer players with powered wheelchairs is involved to get realistic data and scenarios (Handisport Rennes Club). The methodology used in the Interreg ADAPT project will be adapted to settle experimentations with the help of the research team of the rehabilitation center Pôle Saint Hélier. They are qualified to manage experiments with disabled people with a relevant methodology and handle the process to get the agreement to proceed such experiments.
The thesis will be conducted within a multidisciplinary team combining skills in robotics, image analysis, virtual reality, haptics, mechanics and electronics, and will be based on the recommendations and clinical expertise of the Pôle Saint Hélier. The recruited person will be in close connection with a local team of soccer players with powered wheelchairs (Handisport Rennes Club).
(c) GdR 720 ISIS - CNRS - 2011-2022.