Detection of uneven terrains : walking on uneven terrains with a bionic prosthesis for the lower-limb

The technological innovations are becoming more and more present to facilitate everyday life. From the robot vacuum cleaner to the autonomous car, they cover several areas. The one we are interested in is the engineering focused on developing active prosthesis for the amputees of a lower-limb. This can help them in their everyday life, specifically when walking on uneven terrains. Indeed, such terrains cause changes in the human gait. Therefore, we study their impacts on the gait to know which ones are useful to detect, i.e. cobblestones, high grass, meadow, rocks, sand and woodchips. Once this question answered, a method to detect uneven terrains, called convolutional neural network, is built and evaluated on its performances of classification. Based on 6 classifiers, one for each terrain, the conclusion is that confusion often occurs. The aim of this thesis is a first step towards an autonomous prosthesis.