Skip Navigation

ECE Deparmental Seminar 

Robust Statistical Inference in Hierarchically Gaussian Dynamic Systems

Dr. Jordi Vilà-Valls

Centre Tecnològic de Telecomunicacions de Catalunya (CTTC)

Monday, 11/20/17, 11:00am
Light Engineering 250

Abstract: Bayesian inference appears in many signal processing problems, reason why it has attracted the attention of many researchers to develop efficient algorithms, yet computationally affordable. Ranging from KF to SMC methods, there is a plethora of alternatives depending on model assumptions, application requirements, and available computational resources. In many real-life problems, traditional assumptions such as Gaussianity and fully known noise statistics do not hold. In such applications, symmetric and skewed non-Gaussian noise distributions are useful to model possible outliers, impulsive behaviors or the underlying characteristics of the physical phenomena under study, which may be unknown to a certain extent. Regarding the derivation of new robust filtering methods for dynamic complex systems, a problem of interest is how to deal with uncertainties or model mismatch. In this talk we discuss a new robust Bayesian inference approach in hierarchically Gaussian dynamic systems. We show results on representative radar target tracking and indoor localization examples.

Bio: Jordi Vilà-Valls is a researcher at Centre Tecnològic de Telecomunicacions de Catalunya (CTTC) and part-time lecturer at the Telecommunications and Aerospace Engineering School, Universitat Politècnica de Catalunya (EETAC-UPC), Barcelona, Spain. He received his PhD in Electrical Engineering from Grenoble INP, France, in 2010. His primary areas of interest include statistical signal processing at large, estimation and detection theory, nonlinear Bayesian inference, robustness and adaptive methods; with applications to GNSS, indoor positioning/localization, tracking systems, wireless communications and aerospace science.