EMDs with amplitude information for distributed fusion

Abstract

Exponential Mixture Densities (EMDs) is increasingly popular as a suboptimal distributed fusion technique that avoids calculating the common information between different nodes. However, there exists some concerns about the EMDs because it fuses the cluttered posterior density as a whole, which contains plenty of components of little physical significance. Thus, it becomes intractable and computation expensive especially when targets are closely spaced or heavy clutters are distributed in the vicinity of targets. To address this problem, in this paper, a EMDs-based fusion algorithm with amplitude information is proposed. Considering the amplitude of target returns is stronger than that coming from false alarm, and the amplitude from each target is distinctly different, here, the amplitude information is utilized to identify targets and clutters. We implement this approach using Gaussian Mixture techniques and demonstrate the effectiveness and high estimation accuracy of the proposed algorithm over the EMDs algorithm and traditional Covariance Intersection (CI) algorithm.