一种改进的分布式数据Chernoff融合方法

Modified Chernoff Fusion Algorithm for Distributed Data

提出了在实际分布式数据融合问题中可用于未知相关性概率密度函数的广义融合算法。在由于信号存在统计相关性引起谣言传播问题的分布式传感环境中,该算法能进行任何数量的概率密度函数的数据融合。分布式传感系统间的互操作要求限定了系统无法对输入进行预处理以确保统计的独立,而协方差交叉算法和快速协方差交叉算法只适合处理高斯信号等独立的输入信号。在未知相关性概率密度函数的情况下,通过使融合概率密度函数的Chernoff信息最小化,可达到任意数量非高斯输入的融合目的。仿真结果表明,该算法融合效果较好。

A generalized fusion algorithm that can be used for unknown correlation probability density function in practical distributed data fusion problems is proposed. In a distributed sensing environment where rumors may propagate due to statistical correlation of signals, this algorithm can perform data fusion with any number of probability density functions. The interoperability requirements of between the distributed sensing systems exert the limits to the input preprocessing to ensure statistical independence, while the covariance crossover algorithm and the fast covariance crossover algorithm are only suitable for processing independent input signals such as Gaussian signals. In the case of an unknown correlation probability density function, the fusion purpose of any number of non-Gaussian inputs may be achieved by minimizing the Chernoff information of the fusion probability density function. The simulation results indicate that this proposed algorithm has fairly good fusion effect.