From sequences of moving humans represented as meshes, the objective of this project is to analyze these sequences in spectral domain. These analyzes, like identifying body parts or mesh matching between two sequences, are given to a neural network. The aim of this network is to create movements of humans of any shape without filming them. Finally, these new movements will be added to an easilly exploitable atlas.
Main themes of this thesis are : Polygonal meshes, Signal processing, Machine Learning.