Ronsse, RenaudDraguet, CamilleCamilleDraguet2025-05-142025-05-142025-05-142020https://hdl.handle.net/2078.2/17152The recent development of powered lower-limb prostheses allows patients suffering from a lower-limb amputation to participate in daily life activities. However, the use of such a prosthesis requires the user to be able to control it in real time, in order to adapt the device to his locomotion modes. Multiple strategies exist but pattern recognition, in particular, has shown promising results for real-time control. To this end, the subsequent work investigates the implementation of a Dynamic Bayesian Network that is built based on data from a small number of mechanical sensors, worn by two able-bodied subjects. This method aims at distinguishing multiple locomotion modes such as the level-ground walking mode, the standing mode, the stair ascent and descent modes, as well as the ramp ascent and descent modes. The experiments that are exposed in this thesis were carried out on a limited number of subjects. Given the small amount of data collected, the classification results obtained were insufficient. However, promising outcomes are presented in the Discussion chapter and the Dynamic Bayesian Network managed to classify tasks in a very short delay of approximately 0.00145 [s]. Therefore, this Master Thesis paves the way for future studies about locomotion intention detection through the use of a Dynamic Bayesian Network.Dynamic Bayesian NetworksFeature selectionLocomotion detectiontext::thesis::master thesisthesis:25101