分布式检测系统的幂求和融合准则

在基于局部检测统计量的分布式检测系统中,传统的直接求和融合准则在面对局部信噪比差异较大时,检测概率下降明显。在详细分析性能损失的原因之后,扩展出一类新融合准则。新准则通过对局部统计量的幂求和,实现对高信噪比数据的更有效利用。对一个双传感器并行分布式检测系统的仿真表明,高次幂求和在局部信噪比差异明显时,性能好于低次幂,当差异减小时则相反,低次幂求和的性能更好些。高次幂融合准则的鲁棒性更强。面对不同使用环境,应该综合考虑各局部传感器脉冲积累数目和局部信噪比的可能取值范围,从而选择合理的幂阶数进行融合。

Power sum fusion rules for distributed detection system

In the detection with a distributed system based on local statistics,the performance of traditional sum fusion rule is degraded a lot if the change of the local signal to noise ratios(SNRs) is great.After theoretical analysis of this degradation,a novel fusion rule based on local statistics is proposed.In order to use the data with higher SNR more efficiently,the higher powers of the local R statistics are summed up with this rule.Numerical simulation results show that the new method achieves higher detection probability when the local SNRs have significant difference,while the performance is lower a little if the difference is small.The fusion rule based on higher power sum is more robust than the direct sum rule.Other important factors (such as the number of observations in local sensors in each decision and the possible value range of local SNRs) should also be considered carefully to choose a suitable power fusion rule for different cases.