Variational EM and non-convex optimization for multisensor scene analysis — LAAS

Next december 22nd I will be giving a seminar at the Robot-Action-Perception team of LAAS/CNRS, whose abstract reads below.

In this talk I describe the mathematical foundations we used in the recent past to address four different multisensor scene analysis tasks, namely: audio-visual speaker detection and localization, separation of moving sound sources, geometric sound source localization and head/body pose estimation. The methods used to solve these two tasks belong to two big groups: probabilistic methods and non-linear optimization methods with constraints. The talk focuses in the mathematical foundations, from the formulation of the real problem, to the proposed algorithmic solution associated to the chosen model. Results on publicly available datasets will be shown and discussed.

Leave a Reply

Your email address will not be published. Required fields are marked *