Predicting affordances during human locomotion using computer vision by Ali Hussein Abdulzahra AL-DABBAGH

For the degree of Doctor of Engineering Sciences and Technology

The exponential growth of technology opened the door for a new generation of locomotion assistive devices (e.g. prostheses and exoskeletons) to emerge. However, such systems still face several challenges regarding their interactions with human users. For instance, they usually lack the capability to timely detect the actions and intentions of the human wearer. Therefore, transitions between two locomotion modes – such as overground walking and stair climbing – are still handled in a reactive way. This thesis introduces a computer vision-based terrain detection algorithm for predicting locomotion modes. It relies on a depth camera attached to the user’s chest and an algorithm processing the acquired point cloud data to predict the terrain type and estimate its geometrical features for multiple steps in front of the user. Additionally, the system is equipped with an estimator of the walking stride length, speed, and turning angle, based on a single inertial sensor that is integrated in the depth camera. The algorithm is validated with 8 participants in several indoor and outdoor paths with a variety of terrain types. Moreover, validation is also achieved by conducting experiments with clear and partially occluded path conditions. Consequently, the thesis provides a validated software package consisting of terrain detection for multiple steps; estimation of geometrical features, speed, and turning angle; with and without partial occlusion. It can provide a valuable tool for a variety of disabled people including amputees with an active prosthesis, visually impaired, or blind people who face locomotion challenges in daily life.

Jury members :

• Prof. Renaud Ronsse (UCLouvain, Belgium), supervisor
• Prof. Hervé Jeanmart (UCLouvain, Belgium), chairperson
• Prof. Bruno Dehez (UCLouvain, Belgium)
• Prof. Christophe De Vleeschouwer (UCLouvain, Belgium) secrétaire
• Prof. Renaud Detry (UCLouvain, Belgium)
• Prof. Eren Erdal Aksoy (Halmstad University, Sweden)