CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Robust spatial blind source separation Authors:  Mika Sipila - University of Jyvaskyla (Finland)
Klaus Nordhausen - University of Jyvaskyla (Finland)
Sara Taskinen - University of Jyvaskyla (Finland) [presenting]
Abstract: Assume a spatial blind source separation model in which the observed multivariate spatial data is assumed to be a linear mixture of latent stationary spatially uncorrelated random fields. The goal is then to recover an unknown mixing procedure as well as latent uncorrelated random fields. Recently, spatial blind source separation methods that are based on simultaneous diagonalization of two or more scatter matrices were proposed. In case of uncontaminated data such methods are capable of solving the blind source separation problem, but in presence of outlying observations the methods perform poorly. We propose a robust blind source separation method which uses robust global and local scatter matrices based on generalized spatial signs in simultaneous diagonalization. Simulation studies are used to illustrate robustness and efficiency properties of proposed methods in various scenarios.