Survey on Information Fusion of Probability Density Functions

Abstract

The probability density function contains not only the first-order and the second-order statistics, but also the higher-order statistics and more complex feature information. Information fusion on probability density functions for multi-sensor is a challenging problem to be solved in the field of signal processing, especially in the application prospect of multi-sensor multi-scale information fusion for automatic driving and unmanned system, this problem has gradually attracted much attention. How to design fusion criteria and how to form a unified fusion framework are the to-pics that scientists and engineers are committed to solving. Aiming at the fusion of multiple probability density functions from multi-sensor of random variables (vectors), this paper investigates the existing relevant fusion theories and methods, and provides some design rules, criteria, principles and theorems, such as axiomatic method, optimization method and super Bayesian method. The study is expected to provide some directional guide for the effective solution of the fusion problem.

Keywords

• |probability density function|information fusion|axiomatization|pooling function|supra-bayesian|machine learning