Four Statistical Approaches for Multisensor Data Fusion under Non-Gaussian Noise

Abstract

Multisensor data fusion methods for Gaussian noise are widely reported, while fusion approaches for non-Gaussian noise are seldom met in the literature. In this study, four statistical fusion methods are presented for a mixture of Gaussians noise. These four methods are the minimum variance approach, the maximum kurtosis approach, the minimum kurtosis approach, and the minimum mean absolute error approach. Preliminary numerical simulations demonstrate that the maximum kurtosis method shows the worst fusion performance, while the rest three methods shows equivalent better fusion performance.